{
  "best_metric": null,
  "best_model_checkpoint": null,
  "epoch": 4.0,
  "eval_steps": 500,
  "global_step": 5289,
  "is_hyper_param_search": false,
  "is_local_process_zero": true,
  "is_world_process_zero": true,
  "log_history": [
    {
      "epoch": 0.0,
      "learning_rate": 5e-05,
      "loss": 2.4085,
      "step": 1
    },
    {
      "epoch": 0.0,
      "learning_rate": 4.999527231467474e-05,
      "loss": 2.412,
      "step": 2
    },
    {
      "epoch": 0.0,
      "learning_rate": 4.999054462934947e-05,
      "loss": 2.2412,
      "step": 3
    },
    {
      "epoch": 0.0,
      "learning_rate": 4.998581694402421e-05,
      "loss": 2.7767,
      "step": 4
    },
    {
      "epoch": 0.0,
      "learning_rate": 4.9981089258698944e-05,
      "loss": 2.3475,
      "step": 5
    },
    {
      "epoch": 0.0,
      "learning_rate": 4.997636157337368e-05,
      "loss": 2.2902,
      "step": 6
    },
    {
      "epoch": 0.01,
      "learning_rate": 4.9971633888048415e-05,
      "loss": 2.1618,
      "step": 7
    },
    {
      "epoch": 0.01,
      "learning_rate": 4.9966906202723154e-05,
      "loss": 2.0788,
      "step": 8
    },
    {
      "epoch": 0.01,
      "learning_rate": 4.9962178517397886e-05,
      "loss": 2.3022,
      "step": 9
    },
    {
      "epoch": 0.01,
      "learning_rate": 4.9957450832072625e-05,
      "loss": 2.2085,
      "step": 10
    },
    {
      "epoch": 0.01,
      "learning_rate": 4.995272314674736e-05,
      "loss": 2.0357,
      "step": 11
    },
    {
      "epoch": 0.01,
      "learning_rate": 4.995272314674736e-05,
      "loss": 2.1559,
      "step": 12
    },
    {
      "epoch": 0.01,
      "learning_rate": 4.9947995461422096e-05,
      "loss": 2.0379,
      "step": 13
    },
    {
      "epoch": 0.01,
      "learning_rate": 4.994326777609683e-05,
      "loss": 2.0144,
      "step": 14
    },
    {
      "epoch": 0.01,
      "learning_rate": 4.993854009077156e-05,
      "loss": 1.997,
      "step": 15
    },
    {
      "epoch": 0.01,
      "learning_rate": 4.993381240544629e-05,
      "loss": 2.0751,
      "step": 16
    },
    {
      "epoch": 0.01,
      "learning_rate": 4.992908472012103e-05,
      "loss": 1.8244,
      "step": 17
    },
    {
      "epoch": 0.01,
      "learning_rate": 4.992435703479576e-05,
      "loss": 1.9439,
      "step": 18
    },
    {
      "epoch": 0.01,
      "learning_rate": 4.99196293494705e-05,
      "loss": 1.9008,
      "step": 19
    },
    {
      "epoch": 0.02,
      "learning_rate": 4.9914901664145234e-05,
      "loss": 2.1112,
      "step": 20
    },
    {
      "epoch": 0.02,
      "learning_rate": 4.991017397881997e-05,
      "loss": 2.0481,
      "step": 21
    },
    {
      "epoch": 0.02,
      "learning_rate": 4.9905446293494705e-05,
      "loss": 1.9393,
      "step": 22
    },
    {
      "epoch": 0.02,
      "learning_rate": 4.9900718608169443e-05,
      "loss": 1.8458,
      "step": 23
    },
    {
      "epoch": 0.02,
      "learning_rate": 4.9895990922844175e-05,
      "loss": 1.9741,
      "step": 24
    },
    {
      "epoch": 0.02,
      "learning_rate": 4.9891263237518914e-05,
      "loss": 1.9648,
      "step": 25
    },
    {
      "epoch": 0.02,
      "learning_rate": 4.9886535552193646e-05,
      "loss": 1.9738,
      "step": 26
    },
    {
      "epoch": 0.02,
      "learning_rate": 4.9881807866868385e-05,
      "loss": 1.9869,
      "step": 27
    },
    {
      "epoch": 0.02,
      "learning_rate": 4.987708018154312e-05,
      "loss": 1.8824,
      "step": 28
    },
    {
      "epoch": 0.02,
      "learning_rate": 4.9872352496217856e-05,
      "loss": 1.8664,
      "step": 29
    },
    {
      "epoch": 0.02,
      "learning_rate": 4.986762481089259e-05,
      "loss": 1.8399,
      "step": 30
    },
    {
      "epoch": 0.02,
      "learning_rate": 4.986289712556733e-05,
      "loss": 2.119,
      "step": 31
    },
    {
      "epoch": 0.02,
      "learning_rate": 4.985816944024206e-05,
      "loss": 1.9963,
      "step": 32
    },
    {
      "epoch": 0.02,
      "learning_rate": 4.98534417549168e-05,
      "loss": 1.9873,
      "step": 33
    },
    {
      "epoch": 0.03,
      "learning_rate": 4.984871406959153e-05,
      "loss": 2.006,
      "step": 34
    },
    {
      "epoch": 0.03,
      "learning_rate": 4.984398638426627e-05,
      "loss": 1.9115,
      "step": 35
    },
    {
      "epoch": 0.03,
      "learning_rate": 4.9839258698941e-05,
      "loss": 2.0756,
      "step": 36
    },
    {
      "epoch": 0.03,
      "learning_rate": 4.983453101361574e-05,
      "loss": 2.0229,
      "step": 37
    },
    {
      "epoch": 0.03,
      "learning_rate": 4.982980332829047e-05,
      "loss": 2.0575,
      "step": 38
    },
    {
      "epoch": 0.03,
      "learning_rate": 4.982507564296521e-05,
      "loss": 1.9715,
      "step": 39
    },
    {
      "epoch": 0.03,
      "learning_rate": 4.982034795763994e-05,
      "loss": 2.018,
      "step": 40
    },
    {
      "epoch": 0.03,
      "learning_rate": 4.981562027231468e-05,
      "loss": 1.9759,
      "step": 41
    },
    {
      "epoch": 0.03,
      "learning_rate": 4.9810892586989414e-05,
      "loss": 1.9817,
      "step": 42
    },
    {
      "epoch": 0.03,
      "learning_rate": 4.9806164901664146e-05,
      "loss": 2.0391,
      "step": 43
    },
    {
      "epoch": 0.03,
      "learning_rate": 4.980143721633888e-05,
      "loss": 2.0074,
      "step": 44
    },
    {
      "epoch": 0.03,
      "learning_rate": 4.9796709531013616e-05,
      "loss": 1.9212,
      "step": 45
    },
    {
      "epoch": 0.03,
      "learning_rate": 4.979198184568835e-05,
      "loss": 1.8844,
      "step": 46
    },
    {
      "epoch": 0.04,
      "learning_rate": 4.978725416036309e-05,
      "loss": 1.9906,
      "step": 47
    },
    {
      "epoch": 0.04,
      "learning_rate": 4.978252647503782e-05,
      "loss": 2.041,
      "step": 48
    },
    {
      "epoch": 0.04,
      "learning_rate": 4.977779878971256e-05,
      "loss": 1.8663,
      "step": 49
    },
    {
      "epoch": 0.04,
      "learning_rate": 4.977307110438729e-05,
      "loss": 2.0306,
      "step": 50
    },
    {
      "epoch": 0.04,
      "learning_rate": 4.976834341906203e-05,
      "loss": 2.0081,
      "step": 51
    },
    {
      "epoch": 0.04,
      "learning_rate": 4.976361573373676e-05,
      "loss": 1.9292,
      "step": 52
    },
    {
      "epoch": 0.04,
      "learning_rate": 4.97588880484115e-05,
      "loss": 2.118,
      "step": 53
    },
    {
      "epoch": 0.04,
      "learning_rate": 4.975416036308623e-05,
      "loss": 1.9822,
      "step": 54
    },
    {
      "epoch": 0.04,
      "learning_rate": 4.974943267776097e-05,
      "loss": 2.0308,
      "step": 55
    },
    {
      "epoch": 0.04,
      "learning_rate": 4.97447049924357e-05,
      "loss": 1.96,
      "step": 56
    },
    {
      "epoch": 0.04,
      "learning_rate": 4.973997730711044e-05,
      "loss": 1.9603,
      "step": 57
    },
    {
      "epoch": 0.04,
      "learning_rate": 4.9735249621785174e-05,
      "loss": 2.0642,
      "step": 58
    },
    {
      "epoch": 0.04,
      "learning_rate": 4.973052193645991e-05,
      "loss": 1.8739,
      "step": 59
    },
    {
      "epoch": 0.05,
      "learning_rate": 4.9725794251134645e-05,
      "loss": 1.958,
      "step": 60
    },
    {
      "epoch": 0.05,
      "learning_rate": 4.9721066565809384e-05,
      "loss": 1.9945,
      "step": 61
    },
    {
      "epoch": 0.05,
      "learning_rate": 4.9716338880484116e-05,
      "loss": 1.9297,
      "step": 62
    },
    {
      "epoch": 0.05,
      "learning_rate": 4.9711611195158855e-05,
      "loss": 1.9928,
      "step": 63
    },
    {
      "epoch": 0.05,
      "learning_rate": 4.970688350983359e-05,
      "loss": 1.8398,
      "step": 64
    },
    {
      "epoch": 0.05,
      "learning_rate": 4.9702155824508325e-05,
      "loss": 1.998,
      "step": 65
    },
    {
      "epoch": 0.05,
      "learning_rate": 4.9697428139183064e-05,
      "loss": 1.9349,
      "step": 66
    },
    {
      "epoch": 0.05,
      "learning_rate": 4.9692700453857796e-05,
      "loss": 1.9185,
      "step": 67
    },
    {
      "epoch": 0.05,
      "learning_rate": 4.968797276853253e-05,
      "loss": 1.9925,
      "step": 68
    },
    {
      "epoch": 0.05,
      "learning_rate": 4.968324508320726e-05,
      "loss": 2.0546,
      "step": 69
    },
    {
      "epoch": 0.05,
      "learning_rate": 4.9678517397882e-05,
      "loss": 1.9456,
      "step": 70
    },
    {
      "epoch": 0.05,
      "learning_rate": 4.967378971255673e-05,
      "loss": 1.9429,
      "step": 71
    },
    {
      "epoch": 0.05,
      "learning_rate": 4.966906202723147e-05,
      "loss": 2.0146,
      "step": 72
    },
    {
      "epoch": 0.06,
      "learning_rate": 4.96643343419062e-05,
      "loss": 2.2205,
      "step": 73
    },
    {
      "epoch": 0.06,
      "learning_rate": 4.965960665658094e-05,
      "loss": 1.9092,
      "step": 74
    },
    {
      "epoch": 0.06,
      "learning_rate": 4.965487897125567e-05,
      "loss": 1.9265,
      "step": 75
    },
    {
      "epoch": 0.06,
      "learning_rate": 4.965015128593041e-05,
      "loss": 1.8962,
      "step": 76
    },
    {
      "epoch": 0.06,
      "learning_rate": 4.9645423600605144e-05,
      "loss": 2.0029,
      "step": 77
    },
    {
      "epoch": 0.06,
      "learning_rate": 4.964069591527988e-05,
      "loss": 1.976,
      "step": 78
    },
    {
      "epoch": 0.06,
      "learning_rate": 4.9635968229954615e-05,
      "loss": 1.9505,
      "step": 79
    },
    {
      "epoch": 0.06,
      "learning_rate": 4.9631240544629354e-05,
      "loss": 1.9207,
      "step": 80
    },
    {
      "epoch": 0.06,
      "learning_rate": 4.9626512859304086e-05,
      "loss": 2.0111,
      "step": 81
    },
    {
      "epoch": 0.06,
      "learning_rate": 4.9621785173978825e-05,
      "loss": 1.9453,
      "step": 82
    },
    {
      "epoch": 0.06,
      "learning_rate": 4.961705748865356e-05,
      "loss": 1.9826,
      "step": 83
    },
    {
      "epoch": 0.06,
      "learning_rate": 4.9612329803328296e-05,
      "loss": 1.9824,
      "step": 84
    },
    {
      "epoch": 0.06,
      "learning_rate": 4.960760211800303e-05,
      "loss": 1.9638,
      "step": 85
    },
    {
      "epoch": 0.07,
      "learning_rate": 4.9602874432677766e-05,
      "loss": 1.9017,
      "step": 86
    },
    {
      "epoch": 0.07,
      "learning_rate": 4.95981467473525e-05,
      "loss": 2.0813,
      "step": 87
    },
    {
      "epoch": 0.07,
      "learning_rate": 4.959341906202724e-05,
      "loss": 1.9586,
      "step": 88
    },
    {
      "epoch": 0.07,
      "learning_rate": 4.958869137670197e-05,
      "loss": 1.9559,
      "step": 89
    },
    {
      "epoch": 0.07,
      "learning_rate": 4.958396369137671e-05,
      "loss": 1.9739,
      "step": 90
    },
    {
      "epoch": 0.07,
      "learning_rate": 4.957923600605144e-05,
      "loss": 2.0028,
      "step": 91
    },
    {
      "epoch": 0.07,
      "learning_rate": 4.957450832072618e-05,
      "loss": 1.8794,
      "step": 92
    },
    {
      "epoch": 0.07,
      "learning_rate": 4.956978063540091e-05,
      "loss": 1.913,
      "step": 93
    },
    {
      "epoch": 0.07,
      "learning_rate": 4.956505295007565e-05,
      "loss": 2.0655,
      "step": 94
    },
    {
      "epoch": 0.07,
      "learning_rate": 4.956032526475038e-05,
      "loss": 2.069,
      "step": 95
    },
    {
      "epoch": 0.07,
      "learning_rate": 4.9555597579425114e-05,
      "loss": 2.0179,
      "step": 96
    },
    {
      "epoch": 0.07,
      "learning_rate": 4.9550869894099846e-05,
      "loss": 1.9857,
      "step": 97
    },
    {
      "epoch": 0.07,
      "learning_rate": 4.9546142208774585e-05,
      "loss": 2.0019,
      "step": 98
    },
    {
      "epoch": 0.07,
      "learning_rate": 4.954141452344932e-05,
      "loss": 2.005,
      "step": 99
    },
    {
      "epoch": 0.08,
      "learning_rate": 4.9536686838124056e-05,
      "loss": 2.0512,
      "step": 100
    },
    {
      "epoch": 0.08,
      "learning_rate": 4.953195915279879e-05,
      "loss": 1.9384,
      "step": 101
    },
    {
      "epoch": 0.08,
      "learning_rate": 4.952723146747353e-05,
      "loss": 1.9374,
      "step": 102
    },
    {
      "epoch": 0.08,
      "learning_rate": 4.952250378214826e-05,
      "loss": 2.0586,
      "step": 103
    },
    {
      "epoch": 0.08,
      "learning_rate": 4.9517776096823e-05,
      "loss": 1.9806,
      "step": 104
    },
    {
      "epoch": 0.08,
      "learning_rate": 4.951304841149773e-05,
      "loss": 1.99,
      "step": 105
    },
    {
      "epoch": 0.08,
      "learning_rate": 4.950832072617247e-05,
      "loss": 2.0087,
      "step": 106
    },
    {
      "epoch": 0.08,
      "learning_rate": 4.95035930408472e-05,
      "loss": 2.0059,
      "step": 107
    },
    {
      "epoch": 0.08,
      "learning_rate": 4.949886535552194e-05,
      "loss": 1.9861,
      "step": 108
    },
    {
      "epoch": 0.08,
      "learning_rate": 4.949413767019667e-05,
      "loss": 1.9849,
      "step": 109
    },
    {
      "epoch": 0.08,
      "learning_rate": 4.948940998487141e-05,
      "loss": 1.8764,
      "step": 110
    },
    {
      "epoch": 0.08,
      "learning_rate": 4.948468229954614e-05,
      "loss": 2.0079,
      "step": 111
    },
    {
      "epoch": 0.08,
      "learning_rate": 4.947995461422088e-05,
      "loss": 2.006,
      "step": 112
    },
    {
      "epoch": 0.09,
      "learning_rate": 4.947522692889561e-05,
      "loss": 1.946,
      "step": 113
    },
    {
      "epoch": 0.09,
      "learning_rate": 4.947049924357035e-05,
      "loss": 1.9703,
      "step": 114
    },
    {
      "epoch": 0.09,
      "learning_rate": 4.9465771558245084e-05,
      "loss": 1.9149,
      "step": 115
    },
    {
      "epoch": 0.09,
      "learning_rate": 4.946104387291982e-05,
      "loss": 2.0105,
      "step": 116
    },
    {
      "epoch": 0.09,
      "learning_rate": 4.9456316187594555e-05,
      "loss": 2.0989,
      "step": 117
    },
    {
      "epoch": 0.09,
      "learning_rate": 4.9451588502269294e-05,
      "loss": 1.9225,
      "step": 118
    },
    {
      "epoch": 0.09,
      "learning_rate": 4.9446860816944026e-05,
      "loss": 2.0234,
      "step": 119
    },
    {
      "epoch": 0.09,
      "learning_rate": 4.9442133131618765e-05,
      "loss": 2.1204,
      "step": 120
    },
    {
      "epoch": 0.09,
      "learning_rate": 4.94374054462935e-05,
      "loss": 2.0747,
      "step": 121
    },
    {
      "epoch": 0.09,
      "learning_rate": 4.943267776096823e-05,
      "loss": 2.07,
      "step": 122
    },
    {
      "epoch": 0.09,
      "learning_rate": 4.942795007564297e-05,
      "loss": 2.0418,
      "step": 123
    },
    {
      "epoch": 0.09,
      "learning_rate": 4.94232223903177e-05,
      "loss": 1.876,
      "step": 124
    },
    {
      "epoch": 0.09,
      "learning_rate": 4.941849470499244e-05,
      "loss": 2.1391,
      "step": 125
    },
    {
      "epoch": 0.1,
      "learning_rate": 4.941376701966717e-05,
      "loss": 1.9722,
      "step": 126
    },
    {
      "epoch": 0.1,
      "learning_rate": 4.940903933434191e-05,
      "loss": 1.9892,
      "step": 127
    },
    {
      "epoch": 0.1,
      "learning_rate": 4.940431164901664e-05,
      "loss": 2.0291,
      "step": 128
    },
    {
      "epoch": 0.1,
      "learning_rate": 4.939958396369138e-05,
      "loss": 1.8464,
      "step": 129
    },
    {
      "epoch": 0.1,
      "learning_rate": 4.939485627836611e-05,
      "loss": 1.9236,
      "step": 130
    },
    {
      "epoch": 0.1,
      "learning_rate": 4.939012859304085e-05,
      "loss": 1.8781,
      "step": 131
    },
    {
      "epoch": 0.1,
      "learning_rate": 4.9385400907715583e-05,
      "loss": 2.0918,
      "step": 132
    },
    {
      "epoch": 0.1,
      "learning_rate": 4.938067322239032e-05,
      "loss": 2.0827,
      "step": 133
    },
    {
      "epoch": 0.1,
      "learning_rate": 4.9375945537065054e-05,
      "loss": 2.1419,
      "step": 134
    },
    {
      "epoch": 0.1,
      "learning_rate": 4.937121785173979e-05,
      "loss": 2.0146,
      "step": 135
    },
    {
      "epoch": 0.1,
      "learning_rate": 4.9366490166414525e-05,
      "loss": 2.0217,
      "step": 136
    },
    {
      "epoch": 0.1,
      "learning_rate": 4.9361762481089264e-05,
      "loss": 2.138,
      "step": 137
    },
    {
      "epoch": 0.1,
      "learning_rate": 4.9357034795763996e-05,
      "loss": 2.098,
      "step": 138
    },
    {
      "epoch": 0.11,
      "learning_rate": 4.9352307110438735e-05,
      "loss": 2.1919,
      "step": 139
    },
    {
      "epoch": 0.11,
      "learning_rate": 4.934757942511347e-05,
      "loss": 1.9796,
      "step": 140
    },
    {
      "epoch": 0.11,
      "learning_rate": 4.9342851739788206e-05,
      "loss": 2.0104,
      "step": 141
    },
    {
      "epoch": 0.11,
      "learning_rate": 4.933812405446294e-05,
      "loss": 2.0792,
      "step": 142
    },
    {
      "epoch": 0.11,
      "learning_rate": 4.933339636913768e-05,
      "loss": 1.988,
      "step": 143
    },
    {
      "epoch": 0.11,
      "learning_rate": 4.932866868381241e-05,
      "loss": 2.0873,
      "step": 144
    },
    {
      "epoch": 0.11,
      "learning_rate": 4.932394099848715e-05,
      "loss": 2.4981,
      "step": 145
    },
    {
      "epoch": 0.11,
      "learning_rate": 4.931921331316188e-05,
      "loss": 2.1177,
      "step": 146
    },
    {
      "epoch": 0.11,
      "learning_rate": 4.931448562783662e-05,
      "loss": 1.9658,
      "step": 147
    },
    {
      "epoch": 0.11,
      "learning_rate": 4.930975794251135e-05,
      "loss": 2.1982,
      "step": 148
    },
    {
      "epoch": 0.11,
      "learning_rate": 4.930503025718608e-05,
      "loss": 2.0235,
      "step": 149
    },
    {
      "epoch": 0.11,
      "learning_rate": 4.9300302571860815e-05,
      "loss": 1.9822,
      "step": 150
    },
    {
      "epoch": 0.11,
      "learning_rate": 4.9295574886535554e-05,
      "loss": 2.0329,
      "step": 151
    },
    {
      "epoch": 0.11,
      "learning_rate": 4.9290847201210286e-05,
      "loss": 2.0897,
      "step": 152
    },
    {
      "epoch": 0.12,
      "learning_rate": 4.9286119515885024e-05,
      "loss": 2.0437,
      "step": 153
    },
    {
      "epoch": 0.12,
      "learning_rate": 4.9281391830559757e-05,
      "loss": 1.962,
      "step": 154
    },
    {
      "epoch": 0.12,
      "learning_rate": 4.9276664145234495e-05,
      "loss": 2.0581,
      "step": 155
    },
    {
      "epoch": 0.12,
      "learning_rate": 4.927193645990923e-05,
      "loss": 1.8534,
      "step": 156
    },
    {
      "epoch": 0.12,
      "learning_rate": 4.9267208774583966e-05,
      "loss": 2.0379,
      "step": 157
    },
    {
      "epoch": 0.12,
      "learning_rate": 4.92624810892587e-05,
      "loss": 2.0179,
      "step": 158
    },
    {
      "epoch": 0.12,
      "learning_rate": 4.925775340393344e-05,
      "loss": 1.9645,
      "step": 159
    },
    {
      "epoch": 0.12,
      "learning_rate": 4.925302571860817e-05,
      "loss": 2.0155,
      "step": 160
    },
    {
      "epoch": 0.12,
      "learning_rate": 4.924829803328291e-05,
      "loss": 1.9601,
      "step": 161
    },
    {
      "epoch": 0.12,
      "learning_rate": 4.924357034795764e-05,
      "loss": 1.9497,
      "step": 162
    },
    {
      "epoch": 0.12,
      "learning_rate": 4.923884266263238e-05,
      "loss": 2.0488,
      "step": 163
    },
    {
      "epoch": 0.12,
      "learning_rate": 4.923411497730711e-05,
      "loss": 1.9464,
      "step": 164
    },
    {
      "epoch": 0.12,
      "learning_rate": 4.922938729198185e-05,
      "loss": 1.9058,
      "step": 165
    },
    {
      "epoch": 0.13,
      "learning_rate": 4.922465960665658e-05,
      "loss": 2.0025,
      "step": 166
    },
    {
      "epoch": 0.13,
      "learning_rate": 4.921993192133132e-05,
      "loss": 2.0354,
      "step": 167
    },
    {
      "epoch": 0.13,
      "learning_rate": 4.921520423600605e-05,
      "loss": 1.9644,
      "step": 168
    },
    {
      "epoch": 0.13,
      "learning_rate": 4.921047655068079e-05,
      "loss": 2.0813,
      "step": 169
    },
    {
      "epoch": 0.13,
      "learning_rate": 4.9205748865355524e-05,
      "loss": 2.0769,
      "step": 170
    },
    {
      "epoch": 0.13,
      "learning_rate": 4.920102118003026e-05,
      "loss": 2.1057,
      "step": 171
    },
    {
      "epoch": 0.13,
      "learning_rate": 4.9196293494704995e-05,
      "loss": 2.0048,
      "step": 172
    },
    {
      "epoch": 0.13,
      "learning_rate": 4.9191565809379733e-05,
      "loss": 1.9756,
      "step": 173
    },
    {
      "epoch": 0.13,
      "learning_rate": 4.9186838124054465e-05,
      "loss": 1.9664,
      "step": 174
    },
    {
      "epoch": 0.13,
      "learning_rate": 4.91821104387292e-05,
      "loss": 1.8994,
      "step": 175
    },
    {
      "epoch": 0.13,
      "learning_rate": 4.9177382753403936e-05,
      "loss": 1.8496,
      "step": 176
    },
    {
      "epoch": 0.13,
      "learning_rate": 4.917265506807867e-05,
      "loss": 2.1344,
      "step": 177
    },
    {
      "epoch": 0.13,
      "learning_rate": 4.91679273827534e-05,
      "loss": 2.2317,
      "step": 178
    },
    {
      "epoch": 0.14,
      "learning_rate": 4.916319969742814e-05,
      "loss": 2.0201,
      "step": 179
    },
    {
      "epoch": 0.14,
      "learning_rate": 4.915847201210287e-05,
      "loss": 1.9307,
      "step": 180
    },
    {
      "epoch": 0.14,
      "learning_rate": 4.915374432677761e-05,
      "loss": 1.95,
      "step": 181
    },
    {
      "epoch": 0.14,
      "learning_rate": 4.914901664145234e-05,
      "loss": 2.0594,
      "step": 182
    },
    {
      "epoch": 0.14,
      "learning_rate": 4.914428895612708e-05,
      "loss": 1.9919,
      "step": 183
    },
    {
      "epoch": 0.14,
      "learning_rate": 4.913956127080182e-05,
      "loss": 2.0299,
      "step": 184
    },
    {
      "epoch": 0.14,
      "learning_rate": 4.913483358547655e-05,
      "loss": 2.0294,
      "step": 185
    },
    {
      "epoch": 0.14,
      "learning_rate": 4.913010590015129e-05,
      "loss": 2.0271,
      "step": 186
    },
    {
      "epoch": 0.14,
      "learning_rate": 4.912537821482602e-05,
      "loss": 1.9747,
      "step": 187
    },
    {
      "epoch": 0.14,
      "learning_rate": 4.912065052950076e-05,
      "loss": 2.0002,
      "step": 188
    },
    {
      "epoch": 0.14,
      "learning_rate": 4.9115922844175494e-05,
      "loss": 1.9998,
      "step": 189
    },
    {
      "epoch": 0.14,
      "learning_rate": 4.911119515885023e-05,
      "loss": 1.9361,
      "step": 190
    },
    {
      "epoch": 0.14,
      "learning_rate": 4.9106467473524965e-05,
      "loss": 2.0967,
      "step": 191
    },
    {
      "epoch": 0.15,
      "learning_rate": 4.9101739788199704e-05,
      "loss": 2.0795,
      "step": 192
    },
    {
      "epoch": 0.15,
      "learning_rate": 4.9097012102874436e-05,
      "loss": 1.9912,
      "step": 193
    },
    {
      "epoch": 0.15,
      "learning_rate": 4.9092284417549174e-05,
      "loss": 1.9971,
      "step": 194
    },
    {
      "epoch": 0.15,
      "learning_rate": 4.9087556732223906e-05,
      "loss": 2.0406,
      "step": 195
    },
    {
      "epoch": 0.15,
      "learning_rate": 4.9082829046898645e-05,
      "loss": 2.0048,
      "step": 196
    },
    {
      "epoch": 0.15,
      "learning_rate": 4.907810136157338e-05,
      "loss": 1.901,
      "step": 197
    },
    {
      "epoch": 0.15,
      "learning_rate": 4.9073373676248116e-05,
      "loss": 1.9592,
      "step": 198
    },
    {
      "epoch": 0.15,
      "learning_rate": 4.906864599092285e-05,
      "loss": 1.9288,
      "step": 199
    },
    {
      "epoch": 0.15,
      "learning_rate": 4.906391830559759e-05,
      "loss": 1.9349,
      "step": 200
    },
    {
      "epoch": 0.15,
      "learning_rate": 4.905919062027232e-05,
      "loss": 2.0374,
      "step": 201
    },
    {
      "epoch": 0.15,
      "learning_rate": 4.905446293494705e-05,
      "loss": 1.8165,
      "step": 202
    },
    {
      "epoch": 0.15,
      "learning_rate": 4.904973524962178e-05,
      "loss": 2.0718,
      "step": 203
    },
    {
      "epoch": 0.15,
      "learning_rate": 4.904500756429652e-05,
      "loss": 2.0086,
      "step": 204
    },
    {
      "epoch": 0.16,
      "learning_rate": 4.9040279878971254e-05,
      "loss": 1.9726,
      "step": 205
    },
    {
      "epoch": 0.16,
      "learning_rate": 4.903555219364599e-05,
      "loss": 1.9972,
      "step": 206
    },
    {
      "epoch": 0.16,
      "learning_rate": 4.9030824508320725e-05,
      "loss": 2.0363,
      "step": 207
    },
    {
      "epoch": 0.16,
      "learning_rate": 4.9026096822995464e-05,
      "loss": 1.9986,
      "step": 208
    },
    {
      "epoch": 0.16,
      "learning_rate": 4.9021369137670196e-05,
      "loss": 1.8526,
      "step": 209
    },
    {
      "epoch": 0.16,
      "learning_rate": 4.9016641452344935e-05,
      "loss": 1.9877,
      "step": 210
    },
    {
      "epoch": 0.16,
      "learning_rate": 4.901191376701967e-05,
      "loss": 1.9264,
      "step": 211
    },
    {
      "epoch": 0.16,
      "learning_rate": 4.9007186081694406e-05,
      "loss": 2.046,
      "step": 212
    },
    {
      "epoch": 0.16,
      "learning_rate": 4.900245839636914e-05,
      "loss": 2.1843,
      "step": 213
    },
    {
      "epoch": 0.16,
      "learning_rate": 4.8997730711043877e-05,
      "loss": 2.2355,
      "step": 214
    },
    {
      "epoch": 0.16,
      "learning_rate": 4.899300302571861e-05,
      "loss": 2.2393,
      "step": 215
    },
    {
      "epoch": 0.16,
      "learning_rate": 4.898827534039335e-05,
      "loss": 1.9779,
      "step": 216
    },
    {
      "epoch": 0.16,
      "learning_rate": 4.898354765506808e-05,
      "loss": 2.2316,
      "step": 217
    },
    {
      "epoch": 0.16,
      "learning_rate": 4.897881996974282e-05,
      "loss": 2.0092,
      "step": 218
    },
    {
      "epoch": 0.17,
      "learning_rate": 4.897409228441755e-05,
      "loss": 1.9382,
      "step": 219
    },
    {
      "epoch": 0.17,
      "learning_rate": 4.896936459909229e-05,
      "loss": 1.874,
      "step": 220
    },
    {
      "epoch": 0.17,
      "learning_rate": 4.896463691376702e-05,
      "loss": 1.972,
      "step": 221
    },
    {
      "epoch": 0.17,
      "learning_rate": 4.895990922844176e-05,
      "loss": 1.9034,
      "step": 222
    },
    {
      "epoch": 0.17,
      "learning_rate": 4.895518154311649e-05,
      "loss": 2.1396,
      "step": 223
    },
    {
      "epoch": 0.17,
      "learning_rate": 4.895045385779123e-05,
      "loss": 2.0501,
      "step": 224
    },
    {
      "epoch": 0.17,
      "learning_rate": 4.894572617246596e-05,
      "loss": 2.0231,
      "step": 225
    },
    {
      "epoch": 0.17,
      "learning_rate": 4.89409984871407e-05,
      "loss": 1.9964,
      "step": 226
    },
    {
      "epoch": 0.17,
      "learning_rate": 4.8936270801815434e-05,
      "loss": 2.0806,
      "step": 227
    },
    {
      "epoch": 0.17,
      "learning_rate": 4.8931543116490166e-05,
      "loss": 1.907,
      "step": 228
    },
    {
      "epoch": 0.17,
      "learning_rate": 4.8926815431164905e-05,
      "loss": 1.9331,
      "step": 229
    },
    {
      "epoch": 0.17,
      "learning_rate": 4.892208774583964e-05,
      "loss": 2.0212,
      "step": 230
    },
    {
      "epoch": 0.17,
      "learning_rate": 4.891736006051437e-05,
      "loss": 2.1189,
      "step": 231
    },
    {
      "epoch": 0.18,
      "learning_rate": 4.891263237518911e-05,
      "loss": 2.0265,
      "step": 232
    },
    {
      "epoch": 0.18,
      "learning_rate": 4.890790468986384e-05,
      "loss": 1.8529,
      "step": 233
    },
    {
      "epoch": 0.18,
      "learning_rate": 4.890317700453858e-05,
      "loss": 1.9881,
      "step": 234
    },
    {
      "epoch": 0.18,
      "learning_rate": 4.889844931921331e-05,
      "loss": 2.0176,
      "step": 235
    },
    {
      "epoch": 0.18,
      "learning_rate": 4.889372163388805e-05,
      "loss": 2.0067,
      "step": 236
    },
    {
      "epoch": 0.18,
      "learning_rate": 4.888899394856278e-05,
      "loss": 1.9262,
      "step": 237
    },
    {
      "epoch": 0.18,
      "learning_rate": 4.888426626323752e-05,
      "loss": 2.0144,
      "step": 238
    },
    {
      "epoch": 0.18,
      "learning_rate": 4.887953857791225e-05,
      "loss": 2.0123,
      "step": 239
    },
    {
      "epoch": 0.18,
      "learning_rate": 4.887481089258699e-05,
      "loss": 1.9921,
      "step": 240
    },
    {
      "epoch": 0.18,
      "learning_rate": 4.8870083207261723e-05,
      "loss": 1.9742,
      "step": 241
    },
    {
      "epoch": 0.18,
      "learning_rate": 4.886535552193646e-05,
      "loss": 1.9521,
      "step": 242
    },
    {
      "epoch": 0.18,
      "learning_rate": 4.88606278366112e-05,
      "loss": 2.0376,
      "step": 243
    },
    {
      "epoch": 0.18,
      "learning_rate": 4.885590015128593e-05,
      "loss": 1.9731,
      "step": 244
    },
    {
      "epoch": 0.19,
      "learning_rate": 4.885117246596067e-05,
      "loss": 1.9145,
      "step": 245
    },
    {
      "epoch": 0.19,
      "learning_rate": 4.8846444780635404e-05,
      "loss": 1.9798,
      "step": 246
    },
    {
      "epoch": 0.19,
      "learning_rate": 4.884171709531014e-05,
      "loss": 2.11,
      "step": 247
    },
    {
      "epoch": 0.19,
      "learning_rate": 4.8836989409984875e-05,
      "loss": 2.0855,
      "step": 248
    },
    {
      "epoch": 0.19,
      "learning_rate": 4.8832261724659614e-05,
      "loss": 1.9132,
      "step": 249
    },
    {
      "epoch": 0.19,
      "learning_rate": 4.8827534039334346e-05,
      "loss": 2.0821,
      "step": 250
    },
    {
      "epoch": 0.19,
      "learning_rate": 4.8822806354009085e-05,
      "loss": 2.0119,
      "step": 251
    },
    {
      "epoch": 0.19,
      "learning_rate": 4.881807866868382e-05,
      "loss": 2.1225,
      "step": 252
    },
    {
      "epoch": 0.19,
      "learning_rate": 4.8813350983358556e-05,
      "loss": 2.0861,
      "step": 253
    },
    {
      "epoch": 0.19,
      "learning_rate": 4.880862329803329e-05,
      "loss": 2.0075,
      "step": 254
    },
    {
      "epoch": 0.19,
      "learning_rate": 4.880389561270802e-05,
      "loss": 1.933,
      "step": 255
    },
    {
      "epoch": 0.19,
      "learning_rate": 4.879916792738275e-05,
      "loss": 2.0426,
      "step": 256
    },
    {
      "epoch": 0.19,
      "learning_rate": 4.879444024205749e-05,
      "loss": 1.8887,
      "step": 257
    },
    {
      "epoch": 0.2,
      "learning_rate": 4.878971255673222e-05,
      "loss": 2.0988,
      "step": 258
    },
    {
      "epoch": 0.2,
      "learning_rate": 4.878498487140696e-05,
      "loss": 2.1364,
      "step": 259
    },
    {
      "epoch": 0.2,
      "learning_rate": 4.8780257186081694e-05,
      "loss": 1.8695,
      "step": 260
    },
    {
      "epoch": 0.2,
      "learning_rate": 4.877552950075643e-05,
      "loss": 1.8532,
      "step": 261
    },
    {
      "epoch": 0.2,
      "learning_rate": 4.8770801815431165e-05,
      "loss": 2.0797,
      "step": 262
    },
    {
      "epoch": 0.2,
      "learning_rate": 4.87660741301059e-05,
      "loss": 1.8836,
      "step": 263
    },
    {
      "epoch": 0.2,
      "learning_rate": 4.8761346444780635e-05,
      "loss": 1.8538,
      "step": 264
    },
    {
      "epoch": 0.2,
      "learning_rate": 4.8756618759455374e-05,
      "loss": 2.0143,
      "step": 265
    },
    {
      "epoch": 0.2,
      "learning_rate": 4.8751891074130106e-05,
      "loss": 2.0606,
      "step": 266
    },
    {
      "epoch": 0.2,
      "learning_rate": 4.8747163388804845e-05,
      "loss": 1.9734,
      "step": 267
    },
    {
      "epoch": 0.2,
      "learning_rate": 4.874243570347958e-05,
      "loss": 2.149,
      "step": 268
    },
    {
      "epoch": 0.2,
      "learning_rate": 4.8737708018154316e-05,
      "loss": 1.9293,
      "step": 269
    },
    {
      "epoch": 0.2,
      "learning_rate": 4.873298033282905e-05,
      "loss": 2.1234,
      "step": 270
    },
    {
      "epoch": 0.2,
      "learning_rate": 4.872825264750379e-05,
      "loss": 2.1907,
      "step": 271
    },
    {
      "epoch": 0.21,
      "learning_rate": 4.872352496217852e-05,
      "loss": 2.0523,
      "step": 272
    },
    {
      "epoch": 0.21,
      "learning_rate": 4.871879727685326e-05,
      "loss": 2.01,
      "step": 273
    },
    {
      "epoch": 0.21,
      "learning_rate": 4.871406959152799e-05,
      "loss": 2.0091,
      "step": 274
    },
    {
      "epoch": 0.21,
      "learning_rate": 4.870934190620273e-05,
      "loss": 2.2141,
      "step": 275
    },
    {
      "epoch": 0.21,
      "learning_rate": 4.870461422087746e-05,
      "loss": 2.0128,
      "step": 276
    },
    {
      "epoch": 0.21,
      "learning_rate": 4.86998865355522e-05,
      "loss": 2.0808,
      "step": 277
    },
    {
      "epoch": 0.21,
      "learning_rate": 4.869515885022693e-05,
      "loss": 2.0225,
      "step": 278
    },
    {
      "epoch": 0.21,
      "learning_rate": 4.869043116490167e-05,
      "loss": 2.0025,
      "step": 279
    },
    {
      "epoch": 0.21,
      "learning_rate": 4.86857034795764e-05,
      "loss": 2.0771,
      "step": 280
    },
    {
      "epoch": 0.21,
      "learning_rate": 4.8680975794251135e-05,
      "loss": 2.082,
      "step": 281
    },
    {
      "epoch": 0.21,
      "learning_rate": 4.8676248108925873e-05,
      "loss": 2.1156,
      "step": 282
    },
    {
      "epoch": 0.21,
      "learning_rate": 4.8671520423600606e-05,
      "loss": 1.9408,
      "step": 283
    },
    {
      "epoch": 0.21,
      "learning_rate": 4.866679273827534e-05,
      "loss": 1.9722,
      "step": 284
    },
    {
      "epoch": 0.22,
      "learning_rate": 4.8662065052950076e-05,
      "loss": 1.9945,
      "step": 285
    },
    {
      "epoch": 0.22,
      "learning_rate": 4.865733736762481e-05,
      "loss": 2.12,
      "step": 286
    },
    {
      "epoch": 0.22,
      "learning_rate": 4.865260968229955e-05,
      "loss": 2.0671,
      "step": 287
    },
    {
      "epoch": 0.22,
      "learning_rate": 4.864788199697428e-05,
      "loss": 1.7592,
      "step": 288
    },
    {
      "epoch": 0.22,
      "learning_rate": 4.864315431164902e-05,
      "loss": 1.9844,
      "step": 289
    },
    {
      "epoch": 0.22,
      "learning_rate": 4.863842662632375e-05,
      "loss": 2.1166,
      "step": 290
    },
    {
      "epoch": 0.22,
      "learning_rate": 4.863369894099849e-05,
      "loss": 2.0838,
      "step": 291
    },
    {
      "epoch": 0.22,
      "learning_rate": 4.862897125567322e-05,
      "loss": 2.0535,
      "step": 292
    },
    {
      "epoch": 0.22,
      "learning_rate": 4.862424357034796e-05,
      "loss": 2.0544,
      "step": 293
    },
    {
      "epoch": 0.22,
      "learning_rate": 4.861951588502269e-05,
      "loss": 2.0789,
      "step": 294
    },
    {
      "epoch": 0.22,
      "learning_rate": 4.861478819969743e-05,
      "loss": 2.0141,
      "step": 295
    },
    {
      "epoch": 0.22,
      "learning_rate": 4.861006051437216e-05,
      "loss": 2.0381,
      "step": 296
    },
    {
      "epoch": 0.22,
      "learning_rate": 4.86053328290469e-05,
      "loss": 2.0571,
      "step": 297
    },
    {
      "epoch": 0.23,
      "learning_rate": 4.8600605143721634e-05,
      "loss": 2.011,
      "step": 298
    },
    {
      "epoch": 0.23,
      "learning_rate": 4.859587745839637e-05,
      "loss": 1.9378,
      "step": 299
    },
    {
      "epoch": 0.23,
      "learning_rate": 4.8591149773071105e-05,
      "loss": 1.9265,
      "step": 300
    },
    {
      "epoch": 0.23,
      "learning_rate": 4.8586422087745844e-05,
      "loss": 2.143,
      "step": 301
    },
    {
      "epoch": 0.23,
      "learning_rate": 4.8581694402420576e-05,
      "loss": 2.1641,
      "step": 302
    },
    {
      "epoch": 0.23,
      "learning_rate": 4.8576966717095314e-05,
      "loss": 2.0647,
      "step": 303
    },
    {
      "epoch": 0.23,
      "learning_rate": 4.857223903177005e-05,
      "loss": 1.9862,
      "step": 304
    },
    {
      "epoch": 0.23,
      "learning_rate": 4.8567511346444785e-05,
      "loss": 2.0067,
      "step": 305
    },
    {
      "epoch": 0.23,
      "learning_rate": 4.8562783661119524e-05,
      "loss": 1.9842,
      "step": 306
    },
    {
      "epoch": 0.23,
      "learning_rate": 4.8558055975794256e-05,
      "loss": 2.0996,
      "step": 307
    },
    {
      "epoch": 0.23,
      "learning_rate": 4.855332829046899e-05,
      "loss": 1.95,
      "step": 308
    },
    {
      "epoch": 0.23,
      "learning_rate": 4.854860060514372e-05,
      "loss": 2.0952,
      "step": 309
    },
    {
      "epoch": 0.23,
      "learning_rate": 4.854387291981846e-05,
      "loss": 2.0634,
      "step": 310
    },
    {
      "epoch": 0.24,
      "learning_rate": 4.853914523449319e-05,
      "loss": 2.0028,
      "step": 311
    },
    {
      "epoch": 0.24,
      "learning_rate": 4.853441754916793e-05,
      "loss": 2.0324,
      "step": 312
    },
    {
      "epoch": 0.24,
      "learning_rate": 4.852968986384266e-05,
      "loss": 2.0477,
      "step": 313
    },
    {
      "epoch": 0.24,
      "learning_rate": 4.85249621785174e-05,
      "loss": 2.0424,
      "step": 314
    },
    {
      "epoch": 0.24,
      "learning_rate": 4.852023449319213e-05,
      "loss": 2.0053,
      "step": 315
    },
    {
      "epoch": 0.24,
      "learning_rate": 4.851550680786687e-05,
      "loss": 1.9902,
      "step": 316
    },
    {
      "epoch": 0.24,
      "learning_rate": 4.8510779122541604e-05,
      "loss": 1.9086,
      "step": 317
    },
    {
      "epoch": 0.24,
      "learning_rate": 4.850605143721634e-05,
      "loss": 2.1005,
      "step": 318
    },
    {
      "epoch": 0.24,
      "learning_rate": 4.8501323751891075e-05,
      "loss": 2.0058,
      "step": 319
    },
    {
      "epoch": 0.24,
      "learning_rate": 4.8496596066565814e-05,
      "loss": 1.9474,
      "step": 320
    },
    {
      "epoch": 0.24,
      "learning_rate": 4.8491868381240546e-05,
      "loss": 1.98,
      "step": 321
    },
    {
      "epoch": 0.24,
      "learning_rate": 4.8487140695915285e-05,
      "loss": 2.0892,
      "step": 322
    },
    {
      "epoch": 0.24,
      "learning_rate": 4.848241301059002e-05,
      "loss": 2.0518,
      "step": 323
    },
    {
      "epoch": 0.25,
      "learning_rate": 4.8477685325264755e-05,
      "loss": 2.1857,
      "step": 324
    },
    {
      "epoch": 0.25,
      "learning_rate": 4.847295763993949e-05,
      "loss": 2.0719,
      "step": 325
    },
    {
      "epoch": 0.25,
      "learning_rate": 4.8468229954614226e-05,
      "loss": 2.0755,
      "step": 326
    },
    {
      "epoch": 0.25,
      "learning_rate": 4.846350226928896e-05,
      "loss": 1.9904,
      "step": 327
    },
    {
      "epoch": 0.25,
      "learning_rate": 4.84587745839637e-05,
      "loss": 2.0824,
      "step": 328
    },
    {
      "epoch": 0.25,
      "learning_rate": 4.845404689863843e-05,
      "loss": 1.9394,
      "step": 329
    },
    {
      "epoch": 0.25,
      "learning_rate": 4.844931921331317e-05,
      "loss": 2.0317,
      "step": 330
    },
    {
      "epoch": 0.25,
      "learning_rate": 4.84445915279879e-05,
      "loss": 1.9249,
      "step": 331
    },
    {
      "epoch": 0.25,
      "learning_rate": 4.843986384266264e-05,
      "loss": 2.0625,
      "step": 332
    },
    {
      "epoch": 0.25,
      "learning_rate": 4.843513615733737e-05,
      "loss": 2.1139,
      "step": 333
    },
    {
      "epoch": 0.25,
      "learning_rate": 4.84304084720121e-05,
      "loss": 2.025,
      "step": 334
    },
    {
      "epoch": 0.25,
      "learning_rate": 4.842568078668684e-05,
      "loss": 1.828,
      "step": 335
    },
    {
      "epoch": 0.25,
      "learning_rate": 4.8420953101361574e-05,
      "loss": 2.0947,
      "step": 336
    },
    {
      "epoch": 0.25,
      "learning_rate": 4.8416225416036306e-05,
      "loss": 2.0017,
      "step": 337
    },
    {
      "epoch": 0.26,
      "learning_rate": 4.8411497730711045e-05,
      "loss": 2.0254,
      "step": 338
    },
    {
      "epoch": 0.26,
      "learning_rate": 4.840677004538578e-05,
      "loss": 2.0,
      "step": 339
    },
    {
      "epoch": 0.26,
      "learning_rate": 4.8402042360060516e-05,
      "loss": 1.9429,
      "step": 340
    },
    {
      "epoch": 0.26,
      "learning_rate": 4.839731467473525e-05,
      "loss": 2.0546,
      "step": 341
    },
    {
      "epoch": 0.26,
      "learning_rate": 4.839258698940999e-05,
      "loss": 1.9702,
      "step": 342
    },
    {
      "epoch": 0.26,
      "learning_rate": 4.838785930408472e-05,
      "loss": 1.9797,
      "step": 343
    },
    {
      "epoch": 0.26,
      "learning_rate": 4.838313161875946e-05,
      "loss": 1.9039,
      "step": 344
    },
    {
      "epoch": 0.26,
      "learning_rate": 4.837840393343419e-05,
      "loss": 1.9548,
      "step": 345
    },
    {
      "epoch": 0.26,
      "learning_rate": 4.837367624810893e-05,
      "loss": 1.9396,
      "step": 346
    },
    {
      "epoch": 0.26,
      "learning_rate": 4.836894856278366e-05,
      "loss": 2.0506,
      "step": 347
    },
    {
      "epoch": 0.26,
      "learning_rate": 4.83642208774584e-05,
      "loss": 2.0167,
      "step": 348
    },
    {
      "epoch": 0.26,
      "learning_rate": 4.835949319213313e-05,
      "loss": 2.0363,
      "step": 349
    },
    {
      "epoch": 0.26,
      "learning_rate": 4.835476550680787e-05,
      "loss": 1.978,
      "step": 350
    },
    {
      "epoch": 0.27,
      "learning_rate": 4.83500378214826e-05,
      "loss": 2.0332,
      "step": 351
    },
    {
      "epoch": 0.27,
      "learning_rate": 4.834531013615734e-05,
      "loss": 1.9257,
      "step": 352
    },
    {
      "epoch": 0.27,
      "learning_rate": 4.834058245083207e-05,
      "loss": 1.9969,
      "step": 353
    },
    {
      "epoch": 0.27,
      "learning_rate": 4.833585476550681e-05,
      "loss": 1.8301,
      "step": 354
    },
    {
      "epoch": 0.27,
      "learning_rate": 4.8331127080181544e-05,
      "loss": 1.9831,
      "step": 355
    },
    {
      "epoch": 0.27,
      "learning_rate": 4.832639939485628e-05,
      "loss": 2.055,
      "step": 356
    },
    {
      "epoch": 0.27,
      "learning_rate": 4.8321671709531015e-05,
      "loss": 2.0513,
      "step": 357
    },
    {
      "epoch": 0.27,
      "learning_rate": 4.8316944024205754e-05,
      "loss": 2.089,
      "step": 358
    },
    {
      "epoch": 0.27,
      "learning_rate": 4.8312216338880486e-05,
      "loss": 2.0736,
      "step": 359
    },
    {
      "epoch": 0.27,
      "learning_rate": 4.8307488653555225e-05,
      "loss": 2.0489,
      "step": 360
    },
    {
      "epoch": 0.27,
      "learning_rate": 4.830276096822996e-05,
      "loss": 1.8654,
      "step": 361
    },
    {
      "epoch": 0.27,
      "learning_rate": 4.829803328290469e-05,
      "loss": 2.0741,
      "step": 362
    },
    {
      "epoch": 0.27,
      "learning_rate": 4.829330559757943e-05,
      "loss": 1.8542,
      "step": 363
    },
    {
      "epoch": 0.28,
      "learning_rate": 4.828857791225416e-05,
      "loss": 2.027,
      "step": 364
    },
    {
      "epoch": 0.28,
      "learning_rate": 4.82838502269289e-05,
      "loss": 2.035,
      "step": 365
    },
    {
      "epoch": 0.28,
      "learning_rate": 4.827912254160363e-05,
      "loss": 1.9873,
      "step": 366
    },
    {
      "epoch": 0.28,
      "learning_rate": 4.827439485627837e-05,
      "loss": 1.9848,
      "step": 367
    },
    {
      "epoch": 0.28,
      "learning_rate": 4.82696671709531e-05,
      "loss": 1.9613,
      "step": 368
    },
    {
      "epoch": 0.28,
      "learning_rate": 4.826493948562784e-05,
      "loss": 2.0728,
      "step": 369
    },
    {
      "epoch": 0.28,
      "learning_rate": 4.826021180030257e-05,
      "loss": 2.0528,
      "step": 370
    },
    {
      "epoch": 0.28,
      "learning_rate": 4.825548411497731e-05,
      "loss": 1.9249,
      "step": 371
    },
    {
      "epoch": 0.28,
      "learning_rate": 4.825075642965204e-05,
      "loss": 1.9564,
      "step": 372
    },
    {
      "epoch": 0.28,
      "learning_rate": 4.824602874432678e-05,
      "loss": 2.0631,
      "step": 373
    },
    {
      "epoch": 0.28,
      "learning_rate": 4.8241301059001514e-05,
      "loss": 2.0868,
      "step": 374
    },
    {
      "epoch": 0.28,
      "learning_rate": 4.823657337367625e-05,
      "loss": 2.0533,
      "step": 375
    },
    {
      "epoch": 0.28,
      "learning_rate": 4.8231845688350985e-05,
      "loss": 1.9086,
      "step": 376
    },
    {
      "epoch": 0.29,
      "learning_rate": 4.8227118003025724e-05,
      "loss": 2.0189,
      "step": 377
    },
    {
      "epoch": 0.29,
      "learning_rate": 4.8222390317700456e-05,
      "loss": 2.0063,
      "step": 378
    },
    {
      "epoch": 0.29,
      "learning_rate": 4.8217662632375195e-05,
      "loss": 2.0097,
      "step": 379
    },
    {
      "epoch": 0.29,
      "learning_rate": 4.821293494704993e-05,
      "loss": 2.0765,
      "step": 380
    },
    {
      "epoch": 0.29,
      "learning_rate": 4.8208207261724666e-05,
      "loss": 2.1804,
      "step": 381
    },
    {
      "epoch": 0.29,
      "learning_rate": 4.82034795763994e-05,
      "loss": 1.9955,
      "step": 382
    },
    {
      "epoch": 0.29,
      "learning_rate": 4.819875189107414e-05,
      "loss": 1.9851,
      "step": 383
    },
    {
      "epoch": 0.29,
      "learning_rate": 4.819402420574887e-05,
      "loss": 1.969,
      "step": 384
    },
    {
      "epoch": 0.29,
      "learning_rate": 4.818929652042361e-05,
      "loss": 2.124,
      "step": 385
    },
    {
      "epoch": 0.29,
      "learning_rate": 4.818456883509834e-05,
      "loss": 2.0895,
      "step": 386
    },
    {
      "epoch": 0.29,
      "learning_rate": 4.817984114977307e-05,
      "loss": 2.0718,
      "step": 387
    },
    {
      "epoch": 0.29,
      "learning_rate": 4.817511346444781e-05,
      "loss": 1.9524,
      "step": 388
    },
    {
      "epoch": 0.29,
      "learning_rate": 4.817038577912254e-05,
      "loss": 2.1038,
      "step": 389
    },
    {
      "epoch": 0.29,
      "learning_rate": 4.8165658093797275e-05,
      "loss": 1.9198,
      "step": 390
    },
    {
      "epoch": 0.3,
      "learning_rate": 4.8160930408472013e-05,
      "loss": 1.8899,
      "step": 391
    },
    {
      "epoch": 0.3,
      "learning_rate": 4.8156202723146746e-05,
      "loss": 2.0151,
      "step": 392
    },
    {
      "epoch": 0.3,
      "learning_rate": 4.8151475037821484e-05,
      "loss": 1.9639,
      "step": 393
    },
    {
      "epoch": 0.3,
      "learning_rate": 4.8146747352496216e-05,
      "loss": 1.9401,
      "step": 394
    },
    {
      "epoch": 0.3,
      "learning_rate": 4.8142019667170955e-05,
      "loss": 1.9853,
      "step": 395
    },
    {
      "epoch": 0.3,
      "learning_rate": 4.813729198184569e-05,
      "loss": 1.9596,
      "step": 396
    },
    {
      "epoch": 0.3,
      "learning_rate": 4.8132564296520426e-05,
      "loss": 2.0158,
      "step": 397
    },
    {
      "epoch": 0.3,
      "learning_rate": 4.812783661119516e-05,
      "loss": 1.9706,
      "step": 398
    },
    {
      "epoch": 0.3,
      "learning_rate": 4.81231089258699e-05,
      "loss": 1.9797,
      "step": 399
    },
    {
      "epoch": 0.3,
      "learning_rate": 4.811838124054463e-05,
      "loss": 1.8872,
      "step": 400
    },
    {
      "epoch": 0.3,
      "learning_rate": 4.811365355521937e-05,
      "loss": 2.0337,
      "step": 401
    },
    {
      "epoch": 0.3,
      "learning_rate": 4.81089258698941e-05,
      "loss": 1.9557,
      "step": 402
    },
    {
      "epoch": 0.3,
      "learning_rate": 4.810419818456884e-05,
      "loss": 2.0925,
      "step": 403
    },
    {
      "epoch": 0.31,
      "learning_rate": 4.809947049924357e-05,
      "loss": 1.9941,
      "step": 404
    },
    {
      "epoch": 0.31,
      "learning_rate": 4.809474281391831e-05,
      "loss": 1.9971,
      "step": 405
    },
    {
      "epoch": 0.31,
      "learning_rate": 4.809001512859304e-05,
      "loss": 1.9631,
      "step": 406
    },
    {
      "epoch": 0.31,
      "learning_rate": 4.808528744326778e-05,
      "loss": 1.953,
      "step": 407
    },
    {
      "epoch": 0.31,
      "learning_rate": 4.808055975794251e-05,
      "loss": 1.9553,
      "step": 408
    },
    {
      "epoch": 0.31,
      "learning_rate": 4.807583207261725e-05,
      "loss": 2.0301,
      "step": 409
    },
    {
      "epoch": 0.31,
      "learning_rate": 4.8071104387291984e-05,
      "loss": 2.0984,
      "step": 410
    },
    {
      "epoch": 0.31,
      "learning_rate": 4.806637670196672e-05,
      "loss": 2.0085,
      "step": 411
    },
    {
      "epoch": 0.31,
      "learning_rate": 4.8061649016641455e-05,
      "loss": 2.0306,
      "step": 412
    },
    {
      "epoch": 0.31,
      "learning_rate": 4.805692133131619e-05,
      "loss": 1.9533,
      "step": 413
    },
    {
      "epoch": 0.31,
      "learning_rate": 4.8052193645990925e-05,
      "loss": 2.021,
      "step": 414
    },
    {
      "epoch": 0.31,
      "learning_rate": 4.804746596066566e-05,
      "loss": 2.0063,
      "step": 415
    },
    {
      "epoch": 0.31,
      "learning_rate": 4.8042738275340396e-05,
      "loss": 1.9465,
      "step": 416
    },
    {
      "epoch": 0.32,
      "learning_rate": 4.803801059001513e-05,
      "loss": 2.1276,
      "step": 417
    },
    {
      "epoch": 0.32,
      "learning_rate": 4.803328290468986e-05,
      "loss": 2.0381,
      "step": 418
    },
    {
      "epoch": 0.32,
      "learning_rate": 4.80285552193646e-05,
      "loss": 2.0685,
      "step": 419
    },
    {
      "epoch": 0.32,
      "learning_rate": 4.802382753403933e-05,
      "loss": 2.0841,
      "step": 420
    },
    {
      "epoch": 0.32,
      "learning_rate": 4.801909984871407e-05,
      "loss": 1.9812,
      "step": 421
    },
    {
      "epoch": 0.32,
      "learning_rate": 4.80143721633888e-05,
      "loss": 2.1372,
      "step": 422
    },
    {
      "epoch": 0.32,
      "learning_rate": 4.800964447806354e-05,
      "loss": 2.0334,
      "step": 423
    },
    {
      "epoch": 0.32,
      "learning_rate": 4.800491679273828e-05,
      "loss": 2.0253,
      "step": 424
    },
    {
      "epoch": 0.32,
      "learning_rate": 4.800018910741301e-05,
      "loss": 1.9097,
      "step": 425
    },
    {
      "epoch": 0.32,
      "learning_rate": 4.799546142208775e-05,
      "loss": 1.9575,
      "step": 426
    },
    {
      "epoch": 0.32,
      "learning_rate": 4.799073373676248e-05,
      "loss": 2.187,
      "step": 427
    },
    {
      "epoch": 0.32,
      "learning_rate": 4.798600605143722e-05,
      "loss": 2.055,
      "step": 428
    },
    {
      "epoch": 0.32,
      "learning_rate": 4.7981278366111954e-05,
      "loss": 2.0625,
      "step": 429
    },
    {
      "epoch": 0.33,
      "learning_rate": 4.797655068078669e-05,
      "loss": 2.0664,
      "step": 430
    },
    {
      "epoch": 0.33,
      "learning_rate": 4.7971822995461425e-05,
      "loss": 2.0651,
      "step": 431
    },
    {
      "epoch": 0.33,
      "learning_rate": 4.7967095310136163e-05,
      "loss": 2.0696,
      "step": 432
    },
    {
      "epoch": 0.33,
      "learning_rate": 4.7962367624810896e-05,
      "loss": 2.0453,
      "step": 433
    },
    {
      "epoch": 0.33,
      "learning_rate": 4.7957639939485634e-05,
      "loss": 2.0658,
      "step": 434
    },
    {
      "epoch": 0.33,
      "learning_rate": 4.7952912254160366e-05,
      "loss": 2.1195,
      "step": 435
    },
    {
      "epoch": 0.33,
      "learning_rate": 4.7948184568835105e-05,
      "loss": 2.0814,
      "step": 436
    },
    {
      "epoch": 0.33,
      "learning_rate": 4.794345688350984e-05,
      "loss": 2.01,
      "step": 437
    },
    {
      "epoch": 0.33,
      "learning_rate": 4.7938729198184576e-05,
      "loss": 2.0097,
      "step": 438
    },
    {
      "epoch": 0.33,
      "learning_rate": 4.793400151285931e-05,
      "loss": 1.9284,
      "step": 439
    },
    {
      "epoch": 0.33,
      "learning_rate": 4.792927382753404e-05,
      "loss": 1.9519,
      "step": 440
    },
    {
      "epoch": 0.33,
      "learning_rate": 4.792454614220878e-05,
      "loss": 2.0919,
      "step": 441
    },
    {
      "epoch": 0.33,
      "learning_rate": 4.791981845688351e-05,
      "loss": 1.8452,
      "step": 442
    },
    {
      "epoch": 0.34,
      "learning_rate": 4.791509077155824e-05,
      "loss": 1.9984,
      "step": 443
    },
    {
      "epoch": 0.34,
      "learning_rate": 4.791036308623298e-05,
      "loss": 1.9537,
      "step": 444
    },
    {
      "epoch": 0.34,
      "learning_rate": 4.7905635400907714e-05,
      "loss": 2.0628,
      "step": 445
    },
    {
      "epoch": 0.34,
      "learning_rate": 4.790090771558245e-05,
      "loss": 2.0381,
      "step": 446
    },
    {
      "epoch": 0.34,
      "learning_rate": 4.7896180030257185e-05,
      "loss": 1.9629,
      "step": 447
    },
    {
      "epoch": 0.34,
      "learning_rate": 4.7891452344931924e-05,
      "loss": 2.1311,
      "step": 448
    },
    {
      "epoch": 0.34,
      "learning_rate": 4.7886724659606656e-05,
      "loss": 2.0192,
      "step": 449
    },
    {
      "epoch": 0.34,
      "learning_rate": 4.7881996974281395e-05,
      "loss": 2.0428,
      "step": 450
    },
    {
      "epoch": 0.34,
      "learning_rate": 4.787726928895613e-05,
      "loss": 1.9593,
      "step": 451
    },
    {
      "epoch": 0.34,
      "learning_rate": 4.7872541603630866e-05,
      "loss": 2.0618,
      "step": 452
    },
    {
      "epoch": 0.34,
      "learning_rate": 4.78678139183056e-05,
      "loss": 2.0374,
      "step": 453
    },
    {
      "epoch": 0.34,
      "learning_rate": 4.7863086232980337e-05,
      "loss": 1.9091,
      "step": 454
    },
    {
      "epoch": 0.34,
      "learning_rate": 4.785835854765507e-05,
      "loss": 2.0228,
      "step": 455
    },
    {
      "epoch": 0.34,
      "learning_rate": 4.785363086232981e-05,
      "loss": 1.9976,
      "step": 456
    },
    {
      "epoch": 0.35,
      "learning_rate": 4.784890317700454e-05,
      "loss": 2.0183,
      "step": 457
    },
    {
      "epoch": 0.35,
      "learning_rate": 4.784417549167928e-05,
      "loss": 1.9141,
      "step": 458
    },
    {
      "epoch": 0.35,
      "learning_rate": 4.783944780635401e-05,
      "loss": 1.9611,
      "step": 459
    },
    {
      "epoch": 0.35,
      "learning_rate": 4.783472012102875e-05,
      "loss": 2.1428,
      "step": 460
    },
    {
      "epoch": 0.35,
      "learning_rate": 4.782999243570348e-05,
      "loss": 2.0283,
      "step": 461
    },
    {
      "epoch": 0.35,
      "learning_rate": 4.782526475037822e-05,
      "loss": 2.0156,
      "step": 462
    },
    {
      "epoch": 0.35,
      "learning_rate": 4.782053706505295e-05,
      "loss": 2.0963,
      "step": 463
    },
    {
      "epoch": 0.35,
      "learning_rate": 4.781580937972769e-05,
      "loss": 2.0368,
      "step": 464
    },
    {
      "epoch": 0.35,
      "learning_rate": 4.781108169440242e-05,
      "loss": 1.9851,
      "step": 465
    },
    {
      "epoch": 0.35,
      "learning_rate": 4.780635400907716e-05,
      "loss": 2.0841,
      "step": 466
    },
    {
      "epoch": 0.35,
      "learning_rate": 4.7801626323751894e-05,
      "loss": 2.0512,
      "step": 467
    },
    {
      "epoch": 0.35,
      "learning_rate": 4.7796898638426626e-05,
      "loss": 2.1608,
      "step": 468
    },
    {
      "epoch": 0.35,
      "learning_rate": 4.7792170953101365e-05,
      "loss": 2.0457,
      "step": 469
    },
    {
      "epoch": 0.36,
      "learning_rate": 4.77874432677761e-05,
      "loss": 1.9606,
      "step": 470
    },
    {
      "epoch": 0.36,
      "learning_rate": 4.778271558245083e-05,
      "loss": 2.1006,
      "step": 471
    },
    {
      "epoch": 0.36,
      "learning_rate": 4.777798789712557e-05,
      "loss": 1.9714,
      "step": 472
    },
    {
      "epoch": 0.36,
      "learning_rate": 4.77732602118003e-05,
      "loss": 1.9414,
      "step": 473
    },
    {
      "epoch": 0.36,
      "learning_rate": 4.776853252647504e-05,
      "loss": 2.0564,
      "step": 474
    },
    {
      "epoch": 0.36,
      "learning_rate": 4.776380484114977e-05,
      "loss": 1.9741,
      "step": 475
    },
    {
      "epoch": 0.36,
      "learning_rate": 4.775907715582451e-05,
      "loss": 1.9889,
      "step": 476
    },
    {
      "epoch": 0.36,
      "learning_rate": 4.775434947049924e-05,
      "loss": 1.989,
      "step": 477
    },
    {
      "epoch": 0.36,
      "learning_rate": 4.774962178517398e-05,
      "loss": 1.9534,
      "step": 478
    },
    {
      "epoch": 0.36,
      "learning_rate": 4.774489409984871e-05,
      "loss": 2.055,
      "step": 479
    },
    {
      "epoch": 0.36,
      "learning_rate": 4.774016641452345e-05,
      "loss": 2.0618,
      "step": 480
    },
    {
      "epoch": 0.36,
      "learning_rate": 4.7735438729198183e-05,
      "loss": 1.9862,
      "step": 481
    },
    {
      "epoch": 0.36,
      "learning_rate": 4.773071104387292e-05,
      "loss": 2.057,
      "step": 482
    },
    {
      "epoch": 0.37,
      "learning_rate": 4.772598335854766e-05,
      "loss": 2.122,
      "step": 483
    },
    {
      "epoch": 0.37,
      "learning_rate": 4.772125567322239e-05,
      "loss": 1.9343,
      "step": 484
    },
    {
      "epoch": 0.37,
      "learning_rate": 4.771652798789713e-05,
      "loss": 2.0158,
      "step": 485
    },
    {
      "epoch": 0.37,
      "learning_rate": 4.7711800302571864e-05,
      "loss": 2.0633,
      "step": 486
    },
    {
      "epoch": 0.37,
      "learning_rate": 4.77070726172466e-05,
      "loss": 1.9968,
      "step": 487
    },
    {
      "epoch": 0.37,
      "learning_rate": 4.7702344931921335e-05,
      "loss": 1.9506,
      "step": 488
    },
    {
      "epoch": 0.37,
      "learning_rate": 4.7697617246596074e-05,
      "loss": 2.1298,
      "step": 489
    },
    {
      "epoch": 0.37,
      "learning_rate": 4.7692889561270806e-05,
      "loss": 1.9628,
      "step": 490
    },
    {
      "epoch": 0.37,
      "learning_rate": 4.7688161875945545e-05,
      "loss": 1.9442,
      "step": 491
    },
    {
      "epoch": 0.37,
      "learning_rate": 4.768343419062028e-05,
      "loss": 2.0119,
      "step": 492
    },
    {
      "epoch": 0.37,
      "learning_rate": 4.767870650529501e-05,
      "loss": 2.1189,
      "step": 493
    },
    {
      "epoch": 0.37,
      "learning_rate": 4.767397881996975e-05,
      "loss": 1.9944,
      "step": 494
    },
    {
      "epoch": 0.37,
      "learning_rate": 4.766925113464448e-05,
      "loss": 2.0168,
      "step": 495
    },
    {
      "epoch": 0.38,
      "learning_rate": 4.766452344931921e-05,
      "loss": 1.9918,
      "step": 496
    },
    {
      "epoch": 0.38,
      "learning_rate": 4.765979576399395e-05,
      "loss": 2.0893,
      "step": 497
    },
    {
      "epoch": 0.38,
      "learning_rate": 4.765506807866868e-05,
      "loss": 2.078,
      "step": 498
    },
    {
      "epoch": 0.38,
      "learning_rate": 4.765034039334342e-05,
      "loss": 1.9787,
      "step": 499
    },
    {
      "epoch": 0.38,
      "learning_rate": 4.7645612708018154e-05,
      "loss": 2.0261,
      "step": 500
    },
    {
      "epoch": 0.38,
      "learning_rate": 4.764088502269289e-05,
      "loss": 2.0458,
      "step": 501
    },
    {
      "epoch": 0.38,
      "learning_rate": 4.7636157337367624e-05,
      "loss": 2.0946,
      "step": 502
    },
    {
      "epoch": 0.38,
      "learning_rate": 4.763142965204236e-05,
      "loss": 2.0315,
      "step": 503
    },
    {
      "epoch": 0.38,
      "learning_rate": 4.7626701966717095e-05,
      "loss": 2.0422,
      "step": 504
    },
    {
      "epoch": 0.38,
      "learning_rate": 4.7621974281391834e-05,
      "loss": 1.9925,
      "step": 505
    },
    {
      "epoch": 0.38,
      "learning_rate": 4.7617246596066566e-05,
      "loss": 2.0407,
      "step": 506
    },
    {
      "epoch": 0.38,
      "learning_rate": 4.7612518910741305e-05,
      "loss": 2.1064,
      "step": 507
    },
    {
      "epoch": 0.38,
      "learning_rate": 4.760779122541604e-05,
      "loss": 2.1295,
      "step": 508
    },
    {
      "epoch": 0.38,
      "learning_rate": 4.7603063540090776e-05,
      "loss": 1.9745,
      "step": 509
    },
    {
      "epoch": 0.39,
      "learning_rate": 4.759833585476551e-05,
      "loss": 1.8469,
      "step": 510
    },
    {
      "epoch": 0.39,
      "learning_rate": 4.759360816944025e-05,
      "loss": 1.9734,
      "step": 511
    },
    {
      "epoch": 0.39,
      "learning_rate": 4.758888048411498e-05,
      "loss": 1.9218,
      "step": 512
    },
    {
      "epoch": 0.39,
      "learning_rate": 4.758415279878972e-05,
      "loss": 1.9402,
      "step": 513
    },
    {
      "epoch": 0.39,
      "learning_rate": 4.757942511346445e-05,
      "loss": 1.9133,
      "step": 514
    },
    {
      "epoch": 0.39,
      "learning_rate": 4.757469742813919e-05,
      "loss": 1.9429,
      "step": 515
    },
    {
      "epoch": 0.39,
      "learning_rate": 4.756996974281392e-05,
      "loss": 1.8812,
      "step": 516
    },
    {
      "epoch": 0.39,
      "learning_rate": 4.756524205748866e-05,
      "loss": 1.9573,
      "step": 517
    },
    {
      "epoch": 0.39,
      "learning_rate": 4.756051437216339e-05,
      "loss": 2.0669,
      "step": 518
    },
    {
      "epoch": 0.39,
      "learning_rate": 4.755578668683813e-05,
      "loss": 2.0122,
      "step": 519
    },
    {
      "epoch": 0.39,
      "learning_rate": 4.755105900151286e-05,
      "loss": 2.0268,
      "step": 520
    },
    {
      "epoch": 0.39,
      "learning_rate": 4.7546331316187595e-05,
      "loss": 1.991,
      "step": 521
    },
    {
      "epoch": 0.39,
      "learning_rate": 4.7541603630862333e-05,
      "loss": 2.0765,
      "step": 522
    },
    {
      "epoch": 0.4,
      "learning_rate": 4.7536875945537065e-05,
      "loss": 2.0773,
      "step": 523
    },
    {
      "epoch": 0.4,
      "learning_rate": 4.75321482602118e-05,
      "loss": 1.9775,
      "step": 524
    },
    {
      "epoch": 0.4,
      "learning_rate": 4.7527420574886536e-05,
      "loss": 1.9895,
      "step": 525
    },
    {
      "epoch": 0.4,
      "learning_rate": 4.752269288956127e-05,
      "loss": 2.1144,
      "step": 526
    },
    {
      "epoch": 0.4,
      "learning_rate": 4.751796520423601e-05,
      "loss": 2.0288,
      "step": 527
    },
    {
      "epoch": 0.4,
      "learning_rate": 4.751323751891074e-05,
      "loss": 2.0229,
      "step": 528
    },
    {
      "epoch": 0.4,
      "learning_rate": 4.750850983358548e-05,
      "loss": 1.9717,
      "step": 529
    },
    {
      "epoch": 0.4,
      "learning_rate": 4.750378214826021e-05,
      "loss": 2.0122,
      "step": 530
    },
    {
      "epoch": 0.4,
      "learning_rate": 4.749905446293495e-05,
      "loss": 1.9679,
      "step": 531
    },
    {
      "epoch": 0.4,
      "learning_rate": 4.749432677760968e-05,
      "loss": 2.0692,
      "step": 532
    },
    {
      "epoch": 0.4,
      "learning_rate": 4.748959909228442e-05,
      "loss": 2.054,
      "step": 533
    },
    {
      "epoch": 0.4,
      "learning_rate": 4.748487140695915e-05,
      "loss": 2.1148,
      "step": 534
    },
    {
      "epoch": 0.4,
      "learning_rate": 4.748014372163389e-05,
      "loss": 1.9773,
      "step": 535
    },
    {
      "epoch": 0.41,
      "learning_rate": 4.747541603630862e-05,
      "loss": 1.9317,
      "step": 536
    },
    {
      "epoch": 0.41,
      "learning_rate": 4.747068835098336e-05,
      "loss": 1.9418,
      "step": 537
    },
    {
      "epoch": 0.41,
      "learning_rate": 4.7465960665658094e-05,
      "loss": 1.9951,
      "step": 538
    },
    {
      "epoch": 0.41,
      "learning_rate": 4.746123298033283e-05,
      "loss": 1.8869,
      "step": 539
    },
    {
      "epoch": 0.41,
      "learning_rate": 4.7456505295007565e-05,
      "loss": 2.0852,
      "step": 540
    },
    {
      "epoch": 0.41,
      "learning_rate": 4.7451777609682304e-05,
      "loss": 1.9063,
      "step": 541
    },
    {
      "epoch": 0.41,
      "learning_rate": 4.7447049924357036e-05,
      "loss": 2.1507,
      "step": 542
    },
    {
      "epoch": 0.41,
      "learning_rate": 4.7442322239031774e-05,
      "loss": 2.0798,
      "step": 543
    },
    {
      "epoch": 0.41,
      "learning_rate": 4.743759455370651e-05,
      "loss": 1.9669,
      "step": 544
    },
    {
      "epoch": 0.41,
      "learning_rate": 4.7432866868381245e-05,
      "loss": 1.967,
      "step": 545
    },
    {
      "epoch": 0.41,
      "learning_rate": 4.742813918305598e-05,
      "loss": 2.0925,
      "step": 546
    },
    {
      "epoch": 0.41,
      "learning_rate": 4.7423411497730716e-05,
      "loss": 2.0742,
      "step": 547
    },
    {
      "epoch": 0.41,
      "learning_rate": 4.741868381240545e-05,
      "loss": 2.0493,
      "step": 548
    },
    {
      "epoch": 0.42,
      "learning_rate": 4.741395612708018e-05,
      "loss": 1.8355,
      "step": 549
    },
    {
      "epoch": 0.42,
      "learning_rate": 4.740922844175492e-05,
      "loss": 1.9607,
      "step": 550
    },
    {
      "epoch": 0.42,
      "learning_rate": 4.740450075642965e-05,
      "loss": 2.0426,
      "step": 551
    },
    {
      "epoch": 0.42,
      "learning_rate": 4.739977307110439e-05,
      "loss": 1.9696,
      "step": 552
    },
    {
      "epoch": 0.42,
      "learning_rate": 4.739504538577912e-05,
      "loss": 1.9734,
      "step": 553
    },
    {
      "epoch": 0.42,
      "learning_rate": 4.739031770045386e-05,
      "loss": 2.0136,
      "step": 554
    },
    {
      "epoch": 0.42,
      "learning_rate": 4.738559001512859e-05,
      "loss": 1.9956,
      "step": 555
    },
    {
      "epoch": 0.42,
      "learning_rate": 4.738086232980333e-05,
      "loss": 1.9623,
      "step": 556
    },
    {
      "epoch": 0.42,
      "learning_rate": 4.7376134644478064e-05,
      "loss": 2.0878,
      "step": 557
    },
    {
      "epoch": 0.42,
      "learning_rate": 4.73714069591528e-05,
      "loss": 2.0773,
      "step": 558
    },
    {
      "epoch": 0.42,
      "learning_rate": 4.7366679273827535e-05,
      "loss": 1.9323,
      "step": 559
    },
    {
      "epoch": 0.42,
      "learning_rate": 4.7361951588502274e-05,
      "loss": 2.0508,
      "step": 560
    },
    {
      "epoch": 0.42,
      "learning_rate": 4.7357223903177006e-05,
      "loss": 1.8543,
      "step": 561
    },
    {
      "epoch": 0.43,
      "learning_rate": 4.7352496217851745e-05,
      "loss": 2.0394,
      "step": 562
    },
    {
      "epoch": 0.43,
      "learning_rate": 4.7347768532526477e-05,
      "loss": 2.0244,
      "step": 563
    },
    {
      "epoch": 0.43,
      "learning_rate": 4.7343040847201215e-05,
      "loss": 2.1197,
      "step": 564
    },
    {
      "epoch": 0.43,
      "learning_rate": 4.733831316187595e-05,
      "loss": 2.1361,
      "step": 565
    },
    {
      "epoch": 0.43,
      "learning_rate": 4.7333585476550686e-05,
      "loss": 1.9381,
      "step": 566
    },
    {
      "epoch": 0.43,
      "learning_rate": 4.732885779122542e-05,
      "loss": 1.9249,
      "step": 567
    },
    {
      "epoch": 0.43,
      "learning_rate": 4.732413010590016e-05,
      "loss": 2.0023,
      "step": 568
    },
    {
      "epoch": 0.43,
      "learning_rate": 4.731940242057489e-05,
      "loss": 1.9195,
      "step": 569
    },
    {
      "epoch": 0.43,
      "learning_rate": 4.731467473524963e-05,
      "loss": 2.0467,
      "step": 570
    },
    {
      "epoch": 0.43,
      "learning_rate": 4.730994704992436e-05,
      "loss": 1.9787,
      "step": 571
    },
    {
      "epoch": 0.43,
      "learning_rate": 4.73052193645991e-05,
      "loss": 2.1199,
      "step": 572
    },
    {
      "epoch": 0.43,
      "learning_rate": 4.730049167927383e-05,
      "loss": 1.9617,
      "step": 573
    },
    {
      "epoch": 0.43,
      "learning_rate": 4.729576399394856e-05,
      "loss": 1.9769,
      "step": 574
    },
    {
      "epoch": 0.43,
      "learning_rate": 4.72910363086233e-05,
      "loss": 2.093,
      "step": 575
    },
    {
      "epoch": 0.44,
      "learning_rate": 4.7286308623298034e-05,
      "loss": 1.9938,
      "step": 576
    },
    {
      "epoch": 0.44,
      "learning_rate": 4.7281580937972766e-05,
      "loss": 1.8356,
      "step": 577
    },
    {
      "epoch": 0.44,
      "learning_rate": 4.7276853252647505e-05,
      "loss": 1.9721,
      "step": 578
    },
    {
      "epoch": 0.44,
      "learning_rate": 4.727212556732224e-05,
      "loss": 1.9555,
      "step": 579
    },
    {
      "epoch": 0.44,
      "learning_rate": 4.7267397881996976e-05,
      "loss": 2.1106,
      "step": 580
    },
    {
      "epoch": 0.44,
      "learning_rate": 4.726267019667171e-05,
      "loss": 2.0643,
      "step": 581
    },
    {
      "epoch": 0.44,
      "learning_rate": 4.725794251134645e-05,
      "loss": 2.0726,
      "step": 582
    },
    {
      "epoch": 0.44,
      "learning_rate": 4.725321482602118e-05,
      "loss": 2.0121,
      "step": 583
    },
    {
      "epoch": 0.44,
      "learning_rate": 4.724848714069592e-05,
      "loss": 1.9718,
      "step": 584
    },
    {
      "epoch": 0.44,
      "learning_rate": 4.724375945537065e-05,
      "loss": 1.9407,
      "step": 585
    },
    {
      "epoch": 0.44,
      "learning_rate": 4.723903177004539e-05,
      "loss": 2.0019,
      "step": 586
    },
    {
      "epoch": 0.44,
      "learning_rate": 4.723430408472012e-05,
      "loss": 2.0949,
      "step": 587
    },
    {
      "epoch": 0.44,
      "learning_rate": 4.722957639939486e-05,
      "loss": 2.0807,
      "step": 588
    },
    {
      "epoch": 0.45,
      "learning_rate": 4.722484871406959e-05,
      "loss": 2.092,
      "step": 589
    },
    {
      "epoch": 0.45,
      "learning_rate": 4.722012102874433e-05,
      "loss": 2.0374,
      "step": 590
    },
    {
      "epoch": 0.45,
      "learning_rate": 4.721539334341906e-05,
      "loss": 2.0948,
      "step": 591
    },
    {
      "epoch": 0.45,
      "learning_rate": 4.72106656580938e-05,
      "loss": 2.0181,
      "step": 592
    },
    {
      "epoch": 0.45,
      "learning_rate": 4.720593797276853e-05,
      "loss": 2.0157,
      "step": 593
    },
    {
      "epoch": 0.45,
      "learning_rate": 4.720121028744327e-05,
      "loss": 2.0252,
      "step": 594
    },
    {
      "epoch": 0.45,
      "learning_rate": 4.7196482602118004e-05,
      "loss": 1.9758,
      "step": 595
    },
    {
      "epoch": 0.45,
      "learning_rate": 4.719175491679274e-05,
      "loss": 1.8898,
      "step": 596
    },
    {
      "epoch": 0.45,
      "learning_rate": 4.7187027231467475e-05,
      "loss": 2.0704,
      "step": 597
    },
    {
      "epoch": 0.45,
      "learning_rate": 4.7182299546142214e-05,
      "loss": 2.0232,
      "step": 598
    },
    {
      "epoch": 0.45,
      "learning_rate": 4.7177571860816946e-05,
      "loss": 1.8325,
      "step": 599
    },
    {
      "epoch": 0.45,
      "learning_rate": 4.7172844175491685e-05,
      "loss": 1.937,
      "step": 600
    },
    {
      "epoch": 0.45,
      "learning_rate": 4.716811649016642e-05,
      "loss": 2.0705,
      "step": 601
    },
    {
      "epoch": 0.46,
      "learning_rate": 4.716338880484115e-05,
      "loss": 2.0465,
      "step": 602
    },
    {
      "epoch": 0.46,
      "learning_rate": 4.715866111951589e-05,
      "loss": 1.9913,
      "step": 603
    },
    {
      "epoch": 0.46,
      "learning_rate": 4.715393343419062e-05,
      "loss": 2.0207,
      "step": 604
    },
    {
      "epoch": 0.46,
      "learning_rate": 4.714920574886536e-05,
      "loss": 2.1071,
      "step": 605
    },
    {
      "epoch": 0.46,
      "learning_rate": 4.714447806354009e-05,
      "loss": 2.0473,
      "step": 606
    },
    {
      "epoch": 0.46,
      "learning_rate": 4.713975037821483e-05,
      "loss": 2.0446,
      "step": 607
    },
    {
      "epoch": 0.46,
      "learning_rate": 4.713502269288956e-05,
      "loss": 2.1629,
      "step": 608
    },
    {
      "epoch": 0.46,
      "learning_rate": 4.71302950075643e-05,
      "loss": 1.8831,
      "step": 609
    },
    {
      "epoch": 0.46,
      "learning_rate": 4.712556732223903e-05,
      "loss": 2.0923,
      "step": 610
    },
    {
      "epoch": 0.46,
      "learning_rate": 4.712083963691377e-05,
      "loss": 1.9737,
      "step": 611
    },
    {
      "epoch": 0.46,
      "learning_rate": 4.71161119515885e-05,
      "loss": 1.9521,
      "step": 612
    },
    {
      "epoch": 0.46,
      "learning_rate": 4.711138426626324e-05,
      "loss": 1.9362,
      "step": 613
    },
    {
      "epoch": 0.46,
      "learning_rate": 4.7106656580937974e-05,
      "loss": 2.1398,
      "step": 614
    },
    {
      "epoch": 0.47,
      "learning_rate": 4.710192889561271e-05,
      "loss": 1.9738,
      "step": 615
    },
    {
      "epoch": 0.47,
      "learning_rate": 4.7097201210287445e-05,
      "loss": 1.9791,
      "step": 616
    },
    {
      "epoch": 0.47,
      "learning_rate": 4.7092473524962184e-05,
      "loss": 2.079,
      "step": 617
    },
    {
      "epoch": 0.47,
      "learning_rate": 4.7087745839636916e-05,
      "loss": 1.8996,
      "step": 618
    },
    {
      "epoch": 0.47,
      "learning_rate": 4.7083018154311655e-05,
      "loss": 1.8841,
      "step": 619
    },
    {
      "epoch": 0.47,
      "learning_rate": 4.707829046898639e-05,
      "loss": 2.044,
      "step": 620
    },
    {
      "epoch": 0.47,
      "learning_rate": 4.7073562783661126e-05,
      "loss": 2.0999,
      "step": 621
    },
    {
      "epoch": 0.47,
      "learning_rate": 4.706883509833586e-05,
      "loss": 2.0328,
      "step": 622
    },
    {
      "epoch": 0.47,
      "learning_rate": 4.70641074130106e-05,
      "loss": 1.9476,
      "step": 623
    },
    {
      "epoch": 0.47,
      "learning_rate": 4.705937972768533e-05,
      "loss": 2.1309,
      "step": 624
    },
    {
      "epoch": 0.47,
      "learning_rate": 4.705465204236007e-05,
      "loss": 1.9712,
      "step": 625
    },
    {
      "epoch": 0.47,
      "learning_rate": 4.70499243570348e-05,
      "loss": 1.9772,
      "step": 626
    },
    {
      "epoch": 0.47,
      "learning_rate": 4.704519667170953e-05,
      "loss": 2.0459,
      "step": 627
    },
    {
      "epoch": 0.47,
      "learning_rate": 4.704046898638427e-05,
      "loss": 1.9972,
      "step": 628
    },
    {
      "epoch": 0.48,
      "learning_rate": 4.7035741301059e-05,
      "loss": 2.0651,
      "step": 629
    },
    {
      "epoch": 0.48,
      "learning_rate": 4.7031013615733735e-05,
      "loss": 2.0208,
      "step": 630
    },
    {
      "epoch": 0.48,
      "learning_rate": 4.7026285930408473e-05,
      "loss": 2.0012,
      "step": 631
    },
    {
      "epoch": 0.48,
      "learning_rate": 4.7021558245083205e-05,
      "loss": 2.0035,
      "step": 632
    },
    {
      "epoch": 0.48,
      "learning_rate": 4.7016830559757944e-05,
      "loss": 1.9165,
      "step": 633
    },
    {
      "epoch": 0.48,
      "learning_rate": 4.7012102874432676e-05,
      "loss": 2.0447,
      "step": 634
    },
    {
      "epoch": 0.48,
      "learning_rate": 4.7007375189107415e-05,
      "loss": 1.8521,
      "step": 635
    },
    {
      "epoch": 0.48,
      "learning_rate": 4.700264750378215e-05,
      "loss": 2.0587,
      "step": 636
    },
    {
      "epoch": 0.48,
      "learning_rate": 4.6997919818456886e-05,
      "loss": 1.9826,
      "step": 637
    },
    {
      "epoch": 0.48,
      "learning_rate": 4.699319213313162e-05,
      "loss": 1.9142,
      "step": 638
    },
    {
      "epoch": 0.48,
      "learning_rate": 4.698846444780636e-05,
      "loss": 1.9865,
      "step": 639
    },
    {
      "epoch": 0.48,
      "learning_rate": 4.698373676248109e-05,
      "loss": 1.9934,
      "step": 640
    },
    {
      "epoch": 0.48,
      "learning_rate": 4.697900907715583e-05,
      "loss": 2.0886,
      "step": 641
    },
    {
      "epoch": 0.49,
      "learning_rate": 4.697428139183056e-05,
      "loss": 2.0479,
      "step": 642
    },
    {
      "epoch": 0.49,
      "learning_rate": 4.69695537065053e-05,
      "loss": 2.06,
      "step": 643
    },
    {
      "epoch": 0.49,
      "learning_rate": 4.696482602118003e-05,
      "loss": 1.927,
      "step": 644
    },
    {
      "epoch": 0.49,
      "learning_rate": 4.696009833585477e-05,
      "loss": 1.9611,
      "step": 645
    },
    {
      "epoch": 0.49,
      "learning_rate": 4.69553706505295e-05,
      "loss": 1.8159,
      "step": 646
    },
    {
      "epoch": 0.49,
      "learning_rate": 4.695064296520424e-05,
      "loss": 2.0192,
      "step": 647
    },
    {
      "epoch": 0.49,
      "learning_rate": 4.694591527987897e-05,
      "loss": 2.0049,
      "step": 648
    },
    {
      "epoch": 0.49,
      "learning_rate": 4.694118759455371e-05,
      "loss": 1.8767,
      "step": 649
    },
    {
      "epoch": 0.49,
      "learning_rate": 4.6936459909228444e-05,
      "loss": 2.1358,
      "step": 650
    },
    {
      "epoch": 0.49,
      "learning_rate": 4.693173222390318e-05,
      "loss": 2.0347,
      "step": 651
    },
    {
      "epoch": 0.49,
      "learning_rate": 4.6927004538577914e-05,
      "loss": 1.8809,
      "step": 652
    },
    {
      "epoch": 0.49,
      "learning_rate": 4.692227685325265e-05,
      "loss": 1.9794,
      "step": 653
    },
    {
      "epoch": 0.49,
      "learning_rate": 4.6917549167927385e-05,
      "loss": 1.9197,
      "step": 654
    },
    {
      "epoch": 0.5,
      "learning_rate": 4.691282148260212e-05,
      "loss": 2.0027,
      "step": 655
    },
    {
      "epoch": 0.5,
      "learning_rate": 4.6908093797276856e-05,
      "loss": 2.0009,
      "step": 656
    },
    {
      "epoch": 0.5,
      "learning_rate": 4.690336611195159e-05,
      "loss": 2.1214,
      "step": 657
    },
    {
      "epoch": 0.5,
      "learning_rate": 4.689863842662632e-05,
      "loss": 1.9801,
      "step": 658
    },
    {
      "epoch": 0.5,
      "learning_rate": 4.689391074130106e-05,
      "loss": 2.1264,
      "step": 659
    },
    {
      "epoch": 0.5,
      "learning_rate": 4.688918305597579e-05,
      "loss": 2.097,
      "step": 660
    },
    {
      "epoch": 0.5,
      "learning_rate": 4.688445537065053e-05,
      "loss": 1.9603,
      "step": 661
    },
    {
      "epoch": 0.5,
      "learning_rate": 4.687972768532526e-05,
      "loss": 1.9159,
      "step": 662
    },
    {
      "epoch": 0.5,
      "learning_rate": 4.6875e-05,
      "loss": 1.9831,
      "step": 663
    },
    {
      "epoch": 0.5,
      "learning_rate": 4.687027231467474e-05,
      "loss": 1.9399,
      "step": 664
    },
    {
      "epoch": 0.5,
      "learning_rate": 4.686554462934947e-05,
      "loss": 1.9316,
      "step": 665
    },
    {
      "epoch": 0.5,
      "learning_rate": 4.686081694402421e-05,
      "loss": 2.0463,
      "step": 666
    },
    {
      "epoch": 0.5,
      "learning_rate": 4.685608925869894e-05,
      "loss": 2.0582,
      "step": 667
    },
    {
      "epoch": 0.51,
      "learning_rate": 4.685136157337368e-05,
      "loss": 2.0812,
      "step": 668
    },
    {
      "epoch": 0.51,
      "learning_rate": 4.6846633888048414e-05,
      "loss": 1.9375,
      "step": 669
    },
    {
      "epoch": 0.51,
      "learning_rate": 4.684190620272315e-05,
      "loss": 2.0762,
      "step": 670
    },
    {
      "epoch": 0.51,
      "learning_rate": 4.6837178517397885e-05,
      "loss": 2.1082,
      "step": 671
    },
    {
      "epoch": 0.51,
      "learning_rate": 4.6832450832072623e-05,
      "loss": 1.9596,
      "step": 672
    },
    {
      "epoch": 0.51,
      "learning_rate": 4.6827723146747355e-05,
      "loss": 2.0133,
      "step": 673
    },
    {
      "epoch": 0.51,
      "learning_rate": 4.6822995461422094e-05,
      "loss": 2.0128,
      "step": 674
    },
    {
      "epoch": 0.51,
      "learning_rate": 4.6818267776096826e-05,
      "loss": 1.9473,
      "step": 675
    },
    {
      "epoch": 0.51,
      "learning_rate": 4.6813540090771565e-05,
      "loss": 2.0467,
      "step": 676
    },
    {
      "epoch": 0.51,
      "learning_rate": 4.68088124054463e-05,
      "loss": 1.995,
      "step": 677
    },
    {
      "epoch": 0.51,
      "learning_rate": 4.6804084720121036e-05,
      "loss": 1.9448,
      "step": 678
    },
    {
      "epoch": 0.51,
      "learning_rate": 4.679935703479577e-05,
      "loss": 1.9968,
      "step": 679
    },
    {
      "epoch": 0.51,
      "learning_rate": 4.67946293494705e-05,
      "loss": 2.1408,
      "step": 680
    },
    {
      "epoch": 0.52,
      "learning_rate": 4.678990166414524e-05,
      "loss": 1.9973,
      "step": 681
    },
    {
      "epoch": 0.52,
      "learning_rate": 4.678517397881997e-05,
      "loss": 1.9362,
      "step": 682
    },
    {
      "epoch": 0.52,
      "learning_rate": 4.67804462934947e-05,
      "loss": 2.1031,
      "step": 683
    },
    {
      "epoch": 0.52,
      "learning_rate": 4.677571860816944e-05,
      "loss": 2.0301,
      "step": 684
    },
    {
      "epoch": 0.52,
      "learning_rate": 4.6770990922844174e-05,
      "loss": 1.9666,
      "step": 685
    },
    {
      "epoch": 0.52,
      "learning_rate": 4.676626323751891e-05,
      "loss": 2.0277,
      "step": 686
    },
    {
      "epoch": 0.52,
      "learning_rate": 4.6761535552193645e-05,
      "loss": 2.0901,
      "step": 687
    },
    {
      "epoch": 0.52,
      "learning_rate": 4.6756807866868384e-05,
      "loss": 1.8465,
      "step": 688
    },
    {
      "epoch": 0.52,
      "learning_rate": 4.6752080181543116e-05,
      "loss": 1.9978,
      "step": 689
    },
    {
      "epoch": 0.52,
      "learning_rate": 4.6747352496217855e-05,
      "loss": 1.9709,
      "step": 690
    },
    {
      "epoch": 0.52,
      "learning_rate": 4.674262481089259e-05,
      "loss": 2.033,
      "step": 691
    },
    {
      "epoch": 0.52,
      "learning_rate": 4.6737897125567326e-05,
      "loss": 1.9873,
      "step": 692
    },
    {
      "epoch": 0.52,
      "learning_rate": 4.673316944024206e-05,
      "loss": 2.0021,
      "step": 693
    },
    {
      "epoch": 0.52,
      "learning_rate": 4.6728441754916796e-05,
      "loss": 2.0018,
      "step": 694
    },
    {
      "epoch": 0.53,
      "learning_rate": 4.672371406959153e-05,
      "loss": 2.0492,
      "step": 695
    },
    {
      "epoch": 0.53,
      "learning_rate": 4.671898638426627e-05,
      "loss": 2.0383,
      "step": 696
    },
    {
      "epoch": 0.53,
      "learning_rate": 4.6714258698941e-05,
      "loss": 1.9482,
      "step": 697
    },
    {
      "epoch": 0.53,
      "learning_rate": 4.670953101361574e-05,
      "loss": 2.0221,
      "step": 698
    },
    {
      "epoch": 0.53,
      "learning_rate": 4.670480332829047e-05,
      "loss": 2.1207,
      "step": 699
    },
    {
      "epoch": 0.53,
      "learning_rate": 4.670007564296521e-05,
      "loss": 2.0252,
      "step": 700
    },
    {
      "epoch": 0.53,
      "learning_rate": 4.669534795763994e-05,
      "loss": 1.9724,
      "step": 701
    },
    {
      "epoch": 0.53,
      "learning_rate": 4.669062027231468e-05,
      "loss": 2.0711,
      "step": 702
    },
    {
      "epoch": 0.53,
      "learning_rate": 4.668589258698941e-05,
      "loss": 2.1257,
      "step": 703
    },
    {
      "epoch": 0.53,
      "learning_rate": 4.668116490166415e-05,
      "loss": 1.982,
      "step": 704
    },
    {
      "epoch": 0.53,
      "learning_rate": 4.667643721633888e-05,
      "loss": 2.012,
      "step": 705
    },
    {
      "epoch": 0.53,
      "learning_rate": 4.667170953101362e-05,
      "loss": 2.0941,
      "step": 706
    },
    {
      "epoch": 0.53,
      "learning_rate": 4.6666981845688354e-05,
      "loss": 1.9227,
      "step": 707
    },
    {
      "epoch": 0.54,
      "learning_rate": 4.6662254160363086e-05,
      "loss": 1.9917,
      "step": 708
    },
    {
      "epoch": 0.54,
      "learning_rate": 4.6657526475037825e-05,
      "loss": 1.87,
      "step": 709
    },
    {
      "epoch": 0.54,
      "learning_rate": 4.665279878971256e-05,
      "loss": 2.0099,
      "step": 710
    },
    {
      "epoch": 0.54,
      "learning_rate": 4.664807110438729e-05,
      "loss": 2.0871,
      "step": 711
    },
    {
      "epoch": 0.54,
      "learning_rate": 4.664334341906203e-05,
      "loss": 2.1319,
      "step": 712
    },
    {
      "epoch": 0.54,
      "learning_rate": 4.663861573373676e-05,
      "loss": 2.0312,
      "step": 713
    },
    {
      "epoch": 0.54,
      "learning_rate": 4.66338880484115e-05,
      "loss": 1.99,
      "step": 714
    },
    {
      "epoch": 0.54,
      "learning_rate": 4.662916036308623e-05,
      "loss": 1.9487,
      "step": 715
    },
    {
      "epoch": 0.54,
      "learning_rate": 4.662443267776097e-05,
      "loss": 1.9897,
      "step": 716
    },
    {
      "epoch": 0.54,
      "learning_rate": 4.66197049924357e-05,
      "loss": 2.0693,
      "step": 717
    },
    {
      "epoch": 0.54,
      "learning_rate": 4.661497730711044e-05,
      "loss": 2.0483,
      "step": 718
    },
    {
      "epoch": 0.54,
      "learning_rate": 4.661024962178517e-05,
      "loss": 1.8247,
      "step": 719
    },
    {
      "epoch": 0.54,
      "learning_rate": 4.660552193645991e-05,
      "loss": 1.9816,
      "step": 720
    },
    {
      "epoch": 0.55,
      "learning_rate": 4.660079425113464e-05,
      "loss": 2.1522,
      "step": 721
    },
    {
      "epoch": 0.55,
      "learning_rate": 4.659606656580938e-05,
      "loss": 2.1427,
      "step": 722
    },
    {
      "epoch": 0.55,
      "learning_rate": 4.6591338880484114e-05,
      "loss": 1.9673,
      "step": 723
    },
    {
      "epoch": 0.55,
      "learning_rate": 4.658661119515885e-05,
      "loss": 2.126,
      "step": 724
    },
    {
      "epoch": 0.55,
      "learning_rate": 4.658188350983359e-05,
      "loss": 1.9962,
      "step": 725
    },
    {
      "epoch": 0.55,
      "learning_rate": 4.6577155824508324e-05,
      "loss": 1.9532,
      "step": 726
    },
    {
      "epoch": 0.55,
      "learning_rate": 4.657242813918306e-05,
      "loss": 2.0535,
      "step": 727
    },
    {
      "epoch": 0.55,
      "learning_rate": 4.6567700453857795e-05,
      "loss": 2.149,
      "step": 728
    },
    {
      "epoch": 0.55,
      "learning_rate": 4.6562972768532534e-05,
      "loss": 2.1494,
      "step": 729
    },
    {
      "epoch": 0.55,
      "learning_rate": 4.6558245083207266e-05,
      "loss": 2.0947,
      "step": 730
    },
    {
      "epoch": 0.55,
      "learning_rate": 4.6553517397882005e-05,
      "loss": 2.1933,
      "step": 731
    },
    {
      "epoch": 0.55,
      "learning_rate": 4.654878971255674e-05,
      "loss": 1.9575,
      "step": 732
    },
    {
      "epoch": 0.55,
      "learning_rate": 4.654406202723147e-05,
      "loss": 1.9001,
      "step": 733
    },
    {
      "epoch": 0.56,
      "learning_rate": 4.653933434190621e-05,
      "loss": 2.112,
      "step": 734
    },
    {
      "epoch": 0.56,
      "learning_rate": 4.653460665658094e-05,
      "loss": 1.8703,
      "step": 735
    },
    {
      "epoch": 0.56,
      "learning_rate": 4.652987897125567e-05,
      "loss": 1.9642,
      "step": 736
    },
    {
      "epoch": 0.56,
      "learning_rate": 4.652515128593041e-05,
      "loss": 2.1135,
      "step": 737
    },
    {
      "epoch": 0.56,
      "learning_rate": 4.652042360060514e-05,
      "loss": 2.0098,
      "step": 738
    },
    {
      "epoch": 0.56,
      "learning_rate": 4.651569591527988e-05,
      "loss": 1.9651,
      "step": 739
    },
    {
      "epoch": 0.56,
      "learning_rate": 4.6510968229954613e-05,
      "loss": 2.0317,
      "step": 740
    },
    {
      "epoch": 0.56,
      "learning_rate": 4.650624054462935e-05,
      "loss": 1.986,
      "step": 741
    },
    {
      "epoch": 0.56,
      "learning_rate": 4.6501512859304084e-05,
      "loss": 2.1488,
      "step": 742
    },
    {
      "epoch": 0.56,
      "learning_rate": 4.649678517397882e-05,
      "loss": 1.9573,
      "step": 743
    },
    {
      "epoch": 0.56,
      "learning_rate": 4.6492057488653555e-05,
      "loss": 2.0016,
      "step": 744
    },
    {
      "epoch": 0.56,
      "learning_rate": 4.6487329803328294e-05,
      "loss": 1.9205,
      "step": 745
    },
    {
      "epoch": 0.56,
      "learning_rate": 4.6482602118003026e-05,
      "loss": 2.0317,
      "step": 746
    },
    {
      "epoch": 0.56,
      "learning_rate": 4.6477874432677765e-05,
      "loss": 2.1237,
      "step": 747
    },
    {
      "epoch": 0.57,
      "learning_rate": 4.64731467473525e-05,
      "loss": 2.0307,
      "step": 748
    },
    {
      "epoch": 0.57,
      "learning_rate": 4.6468419062027236e-05,
      "loss": 2.0611,
      "step": 749
    },
    {
      "epoch": 0.57,
      "learning_rate": 4.646369137670197e-05,
      "loss": 1.9167,
      "step": 750
    },
    {
      "epoch": 0.57,
      "learning_rate": 4.645896369137671e-05,
      "loss": 1.9563,
      "step": 751
    },
    {
      "epoch": 0.57,
      "learning_rate": 4.645423600605144e-05,
      "loss": 2.008,
      "step": 752
    },
    {
      "epoch": 0.57,
      "learning_rate": 4.644950832072618e-05,
      "loss": 2.1151,
      "step": 753
    },
    {
      "epoch": 0.57,
      "learning_rate": 4.644478063540091e-05,
      "loss": 1.9419,
      "step": 754
    },
    {
      "epoch": 0.57,
      "learning_rate": 4.644005295007565e-05,
      "loss": 2.0756,
      "step": 755
    },
    {
      "epoch": 0.57,
      "learning_rate": 4.643532526475038e-05,
      "loss": 2.1911,
      "step": 756
    },
    {
      "epoch": 0.57,
      "learning_rate": 4.643059757942512e-05,
      "loss": 1.9557,
      "step": 757
    },
    {
      "epoch": 0.57,
      "learning_rate": 4.642586989409985e-05,
      "loss": 2.0832,
      "step": 758
    },
    {
      "epoch": 0.57,
      "learning_rate": 4.642114220877459e-05,
      "loss": 1.9087,
      "step": 759
    },
    {
      "epoch": 0.57,
      "learning_rate": 4.641641452344932e-05,
      "loss": 2.2157,
      "step": 760
    },
    {
      "epoch": 0.58,
      "learning_rate": 4.6411686838124054e-05,
      "loss": 2.0399,
      "step": 761
    },
    {
      "epoch": 0.58,
      "learning_rate": 4.640695915279879e-05,
      "loss": 2.1366,
      "step": 762
    },
    {
      "epoch": 0.58,
      "learning_rate": 4.6402231467473525e-05,
      "loss": 1.9825,
      "step": 763
    },
    {
      "epoch": 0.58,
      "learning_rate": 4.639750378214826e-05,
      "loss": 2.0659,
      "step": 764
    },
    {
      "epoch": 0.58,
      "learning_rate": 4.6392776096822996e-05,
      "loss": 1.8721,
      "step": 765
    },
    {
      "epoch": 0.58,
      "learning_rate": 4.638804841149773e-05,
      "loss": 1.8921,
      "step": 766
    },
    {
      "epoch": 0.58,
      "learning_rate": 4.638332072617247e-05,
      "loss": 1.9436,
      "step": 767
    },
    {
      "epoch": 0.58,
      "learning_rate": 4.63785930408472e-05,
      "loss": 1.9984,
      "step": 768
    },
    {
      "epoch": 0.58,
      "learning_rate": 4.637386535552194e-05,
      "loss": 2.1202,
      "step": 769
    },
    {
      "epoch": 0.58,
      "learning_rate": 4.636913767019667e-05,
      "loss": 2.0339,
      "step": 770
    },
    {
      "epoch": 0.58,
      "learning_rate": 4.636440998487141e-05,
      "loss": 1.9306,
      "step": 771
    },
    {
      "epoch": 0.58,
      "learning_rate": 4.635968229954614e-05,
      "loss": 2.0467,
      "step": 772
    },
    {
      "epoch": 0.58,
      "learning_rate": 4.635495461422088e-05,
      "loss": 1.9731,
      "step": 773
    },
    {
      "epoch": 0.59,
      "learning_rate": 4.635022692889561e-05,
      "loss": 2.0685,
      "step": 774
    },
    {
      "epoch": 0.59,
      "learning_rate": 4.634549924357035e-05,
      "loss": 2.1566,
      "step": 775
    },
    {
      "epoch": 0.59,
      "learning_rate": 4.634077155824508e-05,
      "loss": 1.9014,
      "step": 776
    },
    {
      "epoch": 0.59,
      "learning_rate": 4.633604387291982e-05,
      "loss": 1.977,
      "step": 777
    },
    {
      "epoch": 0.59,
      "learning_rate": 4.6331316187594554e-05,
      "loss": 2.0002,
      "step": 778
    },
    {
      "epoch": 0.59,
      "learning_rate": 4.632658850226929e-05,
      "loss": 1.9837,
      "step": 779
    },
    {
      "epoch": 0.59,
      "learning_rate": 4.6321860816944025e-05,
      "loss": 2.0002,
      "step": 780
    },
    {
      "epoch": 0.59,
      "learning_rate": 4.6317133131618763e-05,
      "loss": 1.8637,
      "step": 781
    },
    {
      "epoch": 0.59,
      "learning_rate": 4.6312405446293495e-05,
      "loss": 2.0607,
      "step": 782
    },
    {
      "epoch": 0.59,
      "learning_rate": 4.6307677760968234e-05,
      "loss": 2.0389,
      "step": 783
    },
    {
      "epoch": 0.59,
      "learning_rate": 4.630295007564297e-05,
      "loss": 1.9727,
      "step": 784
    },
    {
      "epoch": 0.59,
      "learning_rate": 4.6298222390317705e-05,
      "loss": 2.0044,
      "step": 785
    },
    {
      "epoch": 0.59,
      "learning_rate": 4.629349470499244e-05,
      "loss": 2.0802,
      "step": 786
    },
    {
      "epoch": 0.6,
      "learning_rate": 4.6288767019667176e-05,
      "loss": 1.9076,
      "step": 787
    },
    {
      "epoch": 0.6,
      "learning_rate": 4.628403933434191e-05,
      "loss": 2.0007,
      "step": 788
    },
    {
      "epoch": 0.6,
      "learning_rate": 4.627931164901664e-05,
      "loss": 1.953,
      "step": 789
    },
    {
      "epoch": 0.6,
      "learning_rate": 4.627458396369138e-05,
      "loss": 1.9353,
      "step": 790
    },
    {
      "epoch": 0.6,
      "learning_rate": 4.626985627836611e-05,
      "loss": 2.0078,
      "step": 791
    },
    {
      "epoch": 0.6,
      "learning_rate": 4.626512859304085e-05,
      "loss": 2.0739,
      "step": 792
    },
    {
      "epoch": 0.6,
      "learning_rate": 4.626040090771558e-05,
      "loss": 2.0185,
      "step": 793
    },
    {
      "epoch": 0.6,
      "learning_rate": 4.625567322239032e-05,
      "loss": 1.99,
      "step": 794
    },
    {
      "epoch": 0.6,
      "learning_rate": 4.625094553706505e-05,
      "loss": 2.1165,
      "step": 795
    },
    {
      "epoch": 0.6,
      "learning_rate": 4.624621785173979e-05,
      "loss": 1.9798,
      "step": 796
    },
    {
      "epoch": 0.6,
      "learning_rate": 4.6241490166414524e-05,
      "loss": 2.0058,
      "step": 797
    },
    {
      "epoch": 0.6,
      "learning_rate": 4.623676248108926e-05,
      "loss": 2.213,
      "step": 798
    },
    {
      "epoch": 0.6,
      "learning_rate": 4.6232034795763995e-05,
      "loss": 1.9291,
      "step": 799
    },
    {
      "epoch": 0.61,
      "learning_rate": 4.6227307110438734e-05,
      "loss": 2.013,
      "step": 800
    },
    {
      "epoch": 0.61,
      "learning_rate": 4.6222579425113466e-05,
      "loss": 1.9955,
      "step": 801
    },
    {
      "epoch": 0.61,
      "learning_rate": 4.6217851739788204e-05,
      "loss": 2.1379,
      "step": 802
    },
    {
      "epoch": 0.61,
      "learning_rate": 4.6213124054462937e-05,
      "loss": 2.1353,
      "step": 803
    },
    {
      "epoch": 0.61,
      "learning_rate": 4.6208396369137675e-05,
      "loss": 1.9637,
      "step": 804
    },
    {
      "epoch": 0.61,
      "learning_rate": 4.620366868381241e-05,
      "loss": 1.8686,
      "step": 805
    },
    {
      "epoch": 0.61,
      "learning_rate": 4.6198940998487146e-05,
      "loss": 2.0922,
      "step": 806
    },
    {
      "epoch": 0.61,
      "learning_rate": 4.619421331316188e-05,
      "loss": 1.9728,
      "step": 807
    },
    {
      "epoch": 0.61,
      "learning_rate": 4.618948562783662e-05,
      "loss": 1.8828,
      "step": 808
    },
    {
      "epoch": 0.61,
      "learning_rate": 4.618475794251135e-05,
      "loss": 1.9962,
      "step": 809
    },
    {
      "epoch": 0.61,
      "learning_rate": 4.618003025718609e-05,
      "loss": 2.0978,
      "step": 810
    },
    {
      "epoch": 0.61,
      "learning_rate": 4.617530257186082e-05,
      "loss": 2.0119,
      "step": 811
    },
    {
      "epoch": 0.61,
      "learning_rate": 4.617057488653556e-05,
      "loss": 1.9771,
      "step": 812
    },
    {
      "epoch": 0.61,
      "learning_rate": 4.616584720121029e-05,
      "loss": 2.0655,
      "step": 813
    },
    {
      "epoch": 0.62,
      "learning_rate": 4.616111951588502e-05,
      "loss": 2.072,
      "step": 814
    },
    {
      "epoch": 0.62,
      "learning_rate": 4.615639183055976e-05,
      "loss": 2.0743,
      "step": 815
    },
    {
      "epoch": 0.62,
      "learning_rate": 4.6151664145234494e-05,
      "loss": 2.1449,
      "step": 816
    },
    {
      "epoch": 0.62,
      "learning_rate": 4.6146936459909226e-05,
      "loss": 2.1067,
      "step": 817
    },
    {
      "epoch": 0.62,
      "learning_rate": 4.6142208774583965e-05,
      "loss": 2.0394,
      "step": 818
    },
    {
      "epoch": 0.62,
      "learning_rate": 4.61374810892587e-05,
      "loss": 1.9803,
      "step": 819
    },
    {
      "epoch": 0.62,
      "learning_rate": 4.6132753403933436e-05,
      "loss": 2.1195,
      "step": 820
    },
    {
      "epoch": 0.62,
      "learning_rate": 4.612802571860817e-05,
      "loss": 2.1268,
      "step": 821
    },
    {
      "epoch": 0.62,
      "learning_rate": 4.612329803328291e-05,
      "loss": 2.0395,
      "step": 822
    },
    {
      "epoch": 0.62,
      "learning_rate": 4.611857034795764e-05,
      "loss": 2.0212,
      "step": 823
    },
    {
      "epoch": 0.62,
      "learning_rate": 4.611384266263238e-05,
      "loss": 2.0394,
      "step": 824
    },
    {
      "epoch": 0.62,
      "learning_rate": 4.610911497730711e-05,
      "loss": 2.007,
      "step": 825
    },
    {
      "epoch": 0.62,
      "learning_rate": 4.610438729198185e-05,
      "loss": 2.3236,
      "step": 826
    },
    {
      "epoch": 0.63,
      "learning_rate": 4.609965960665658e-05,
      "loss": 2.1389,
      "step": 827
    },
    {
      "epoch": 0.63,
      "learning_rate": 4.609493192133132e-05,
      "loss": 2.0798,
      "step": 828
    },
    {
      "epoch": 0.63,
      "learning_rate": 4.609020423600605e-05,
      "loss": 2.0415,
      "step": 829
    },
    {
      "epoch": 0.63,
      "learning_rate": 4.608547655068079e-05,
      "loss": 2.0602,
      "step": 830
    },
    {
      "epoch": 0.63,
      "learning_rate": 4.608074886535552e-05,
      "loss": 1.9149,
      "step": 831
    },
    {
      "epoch": 0.63,
      "learning_rate": 4.607602118003026e-05,
      "loss": 1.9145,
      "step": 832
    },
    {
      "epoch": 0.63,
      "learning_rate": 4.607129349470499e-05,
      "loss": 2.1503,
      "step": 833
    },
    {
      "epoch": 0.63,
      "learning_rate": 4.606656580937973e-05,
      "loss": 1.977,
      "step": 834
    },
    {
      "epoch": 0.63,
      "learning_rate": 4.6061838124054464e-05,
      "loss": 2.0126,
      "step": 835
    },
    {
      "epoch": 0.63,
      "learning_rate": 4.60571104387292e-05,
      "loss": 2.0527,
      "step": 836
    },
    {
      "epoch": 0.63,
      "learning_rate": 4.6052382753403935e-05,
      "loss": 1.8421,
      "step": 837
    },
    {
      "epoch": 0.63,
      "learning_rate": 4.6047655068078674e-05,
      "loss": 1.9348,
      "step": 838
    },
    {
      "epoch": 0.63,
      "learning_rate": 4.6042927382753406e-05,
      "loss": 2.0096,
      "step": 839
    },
    {
      "epoch": 0.64,
      "learning_rate": 4.6038199697428145e-05,
      "loss": 2.0469,
      "step": 840
    },
    {
      "epoch": 0.64,
      "learning_rate": 4.603347201210288e-05,
      "loss": 1.9703,
      "step": 841
    },
    {
      "epoch": 0.64,
      "learning_rate": 4.602874432677761e-05,
      "loss": 1.7962,
      "step": 842
    },
    {
      "epoch": 0.64,
      "learning_rate": 4.602401664145235e-05,
      "loss": 2.1118,
      "step": 843
    },
    {
      "epoch": 0.64,
      "learning_rate": 4.601928895612708e-05,
      "loss": 2.0867,
      "step": 844
    },
    {
      "epoch": 0.64,
      "learning_rate": 4.601456127080182e-05,
      "loss": 2.1095,
      "step": 845
    },
    {
      "epoch": 0.64,
      "learning_rate": 4.600983358547655e-05,
      "loss": 1.8726,
      "step": 846
    },
    {
      "epoch": 0.64,
      "learning_rate": 4.600510590015129e-05,
      "loss": 2.0879,
      "step": 847
    },
    {
      "epoch": 0.64,
      "learning_rate": 4.600037821482602e-05,
      "loss": 2.0028,
      "step": 848
    },
    {
      "epoch": 0.64,
      "learning_rate": 4.599565052950076e-05,
      "loss": 1.9042,
      "step": 849
    },
    {
      "epoch": 0.64,
      "learning_rate": 4.599092284417549e-05,
      "loss": 2.0401,
      "step": 850
    },
    {
      "epoch": 0.64,
      "learning_rate": 4.598619515885023e-05,
      "loss": 2.0293,
      "step": 851
    },
    {
      "epoch": 0.64,
      "learning_rate": 4.598146747352496e-05,
      "loss": 1.9845,
      "step": 852
    },
    {
      "epoch": 0.65,
      "learning_rate": 4.59767397881997e-05,
      "loss": 1.9064,
      "step": 853
    },
    {
      "epoch": 0.65,
      "learning_rate": 4.5972012102874434e-05,
      "loss": 2.0855,
      "step": 854
    },
    {
      "epoch": 0.65,
      "learning_rate": 4.596728441754917e-05,
      "loss": 1.9974,
      "step": 855
    },
    {
      "epoch": 0.65,
      "learning_rate": 4.5962556732223905e-05,
      "loss": 1.9134,
      "step": 856
    },
    {
      "epoch": 0.65,
      "learning_rate": 4.5957829046898644e-05,
      "loss": 2.0399,
      "step": 857
    },
    {
      "epoch": 0.65,
      "learning_rate": 4.5953101361573376e-05,
      "loss": 1.9572,
      "step": 858
    },
    {
      "epoch": 0.65,
      "learning_rate": 4.5948373676248115e-05,
      "loss": 1.9667,
      "step": 859
    },
    {
      "epoch": 0.65,
      "learning_rate": 4.594364599092285e-05,
      "loss": 2.1055,
      "step": 860
    },
    {
      "epoch": 0.65,
      "learning_rate": 4.5938918305597586e-05,
      "loss": 2.0484,
      "step": 861
    },
    {
      "epoch": 0.65,
      "learning_rate": 4.593419062027232e-05,
      "loss": 2.0998,
      "step": 862
    },
    {
      "epoch": 0.65,
      "learning_rate": 4.5929462934947057e-05,
      "loss": 1.941,
      "step": 863
    },
    {
      "epoch": 0.65,
      "learning_rate": 4.592473524962179e-05,
      "loss": 1.8648,
      "step": 864
    },
    {
      "epoch": 0.65,
      "learning_rate": 4.592000756429653e-05,
      "loss": 1.9865,
      "step": 865
    },
    {
      "epoch": 0.65,
      "learning_rate": 4.591527987897126e-05,
      "loss": 2.0504,
      "step": 866
    },
    {
      "epoch": 0.66,
      "learning_rate": 4.591055219364599e-05,
      "loss": 2.0295,
      "step": 867
    },
    {
      "epoch": 0.66,
      "learning_rate": 4.590582450832073e-05,
      "loss": 1.9713,
      "step": 868
    },
    {
      "epoch": 0.66,
      "learning_rate": 4.590109682299546e-05,
      "loss": 2.04,
      "step": 869
    },
    {
      "epoch": 0.66,
      "learning_rate": 4.5896369137670195e-05,
      "loss": 2.0269,
      "step": 870
    },
    {
      "epoch": 0.66,
      "learning_rate": 4.589164145234493e-05,
      "loss": 2.0053,
      "step": 871
    },
    {
      "epoch": 0.66,
      "learning_rate": 4.5886913767019665e-05,
      "loss": 1.9942,
      "step": 872
    },
    {
      "epoch": 0.66,
      "learning_rate": 4.5882186081694404e-05,
      "loss": 2.0882,
      "step": 873
    },
    {
      "epoch": 0.66,
      "learning_rate": 4.5877458396369136e-05,
      "loss": 2.0002,
      "step": 874
    },
    {
      "epoch": 0.66,
      "learning_rate": 4.5872730711043875e-05,
      "loss": 2.085,
      "step": 875
    },
    {
      "epoch": 0.66,
      "learning_rate": 4.586800302571861e-05,
      "loss": 1.9907,
      "step": 876
    },
    {
      "epoch": 0.66,
      "learning_rate": 4.5863275340393346e-05,
      "loss": 1.9498,
      "step": 877
    },
    {
      "epoch": 0.66,
      "learning_rate": 4.585854765506808e-05,
      "loss": 1.9665,
      "step": 878
    },
    {
      "epoch": 0.66,
      "learning_rate": 4.585381996974282e-05,
      "loss": 1.972,
      "step": 879
    },
    {
      "epoch": 0.67,
      "learning_rate": 4.584909228441755e-05,
      "loss": 2.0074,
      "step": 880
    },
    {
      "epoch": 0.67,
      "learning_rate": 4.584436459909229e-05,
      "loss": 2.1419,
      "step": 881
    },
    {
      "epoch": 0.67,
      "learning_rate": 4.583963691376702e-05,
      "loss": 1.9611,
      "step": 882
    },
    {
      "epoch": 0.67,
      "learning_rate": 4.583490922844176e-05,
      "loss": 2.0223,
      "step": 883
    },
    {
      "epoch": 0.67,
      "learning_rate": 4.583018154311649e-05,
      "loss": 2.0377,
      "step": 884
    },
    {
      "epoch": 0.67,
      "learning_rate": 4.582545385779123e-05,
      "loss": 1.9962,
      "step": 885
    },
    {
      "epoch": 0.67,
      "learning_rate": 4.582072617246596e-05,
      "loss": 2.056,
      "step": 886
    },
    {
      "epoch": 0.67,
      "learning_rate": 4.58159984871407e-05,
      "loss": 2.0463,
      "step": 887
    },
    {
      "epoch": 0.67,
      "learning_rate": 4.581127080181543e-05,
      "loss": 2.0105,
      "step": 888
    },
    {
      "epoch": 0.67,
      "learning_rate": 4.580654311649017e-05,
      "loss": 2.0342,
      "step": 889
    },
    {
      "epoch": 0.67,
      "learning_rate": 4.5801815431164903e-05,
      "loss": 1.8918,
      "step": 890
    },
    {
      "epoch": 0.67,
      "learning_rate": 4.579708774583964e-05,
      "loss": 2.045,
      "step": 891
    },
    {
      "epoch": 0.67,
      "learning_rate": 4.5792360060514374e-05,
      "loss": 2.1161,
      "step": 892
    },
    {
      "epoch": 0.68,
      "learning_rate": 4.578763237518911e-05,
      "loss": 2.0761,
      "step": 893
    },
    {
      "epoch": 0.68,
      "learning_rate": 4.5782904689863845e-05,
      "loss": 2.1443,
      "step": 894
    },
    {
      "epoch": 0.68,
      "learning_rate": 4.577817700453858e-05,
      "loss": 2.1203,
      "step": 895
    },
    {
      "epoch": 0.68,
      "learning_rate": 4.5773449319213316e-05,
      "loss": 1.9773,
      "step": 896
    },
    {
      "epoch": 0.68,
      "learning_rate": 4.576872163388805e-05,
      "loss": 1.9783,
      "step": 897
    },
    {
      "epoch": 0.68,
      "learning_rate": 4.576399394856278e-05,
      "loss": 1.9718,
      "step": 898
    },
    {
      "epoch": 0.68,
      "learning_rate": 4.575926626323752e-05,
      "loss": 1.999,
      "step": 899
    },
    {
      "epoch": 0.68,
      "learning_rate": 4.575453857791225e-05,
      "loss": 2.143,
      "step": 900
    },
    {
      "epoch": 0.68,
      "learning_rate": 4.574981089258699e-05,
      "loss": 2.0757,
      "step": 901
    },
    {
      "epoch": 0.68,
      "learning_rate": 4.574508320726172e-05,
      "loss": 2.0495,
      "step": 902
    },
    {
      "epoch": 0.68,
      "learning_rate": 4.574035552193646e-05,
      "loss": 1.8876,
      "step": 903
    },
    {
      "epoch": 0.68,
      "learning_rate": 4.57356278366112e-05,
      "loss": 2.1422,
      "step": 904
    },
    {
      "epoch": 0.68,
      "learning_rate": 4.573090015128593e-05,
      "loss": 2.0562,
      "step": 905
    },
    {
      "epoch": 0.69,
      "learning_rate": 4.572617246596067e-05,
      "loss": 1.9954,
      "step": 906
    },
    {
      "epoch": 0.69,
      "learning_rate": 4.57214447806354e-05,
      "loss": 2.0707,
      "step": 907
    },
    {
      "epoch": 0.69,
      "learning_rate": 4.571671709531014e-05,
      "loss": 1.9196,
      "step": 908
    },
    {
      "epoch": 0.69,
      "learning_rate": 4.5711989409984874e-05,
      "loss": 1.9723,
      "step": 909
    },
    {
      "epoch": 0.69,
      "learning_rate": 4.570726172465961e-05,
      "loss": 1.9629,
      "step": 910
    },
    {
      "epoch": 0.69,
      "learning_rate": 4.5702534039334344e-05,
      "loss": 1.995,
      "step": 911
    },
    {
      "epoch": 0.69,
      "learning_rate": 4.569780635400908e-05,
      "loss": 2.0749,
      "step": 912
    },
    {
      "epoch": 0.69,
      "learning_rate": 4.5693078668683815e-05,
      "loss": 2.0445,
      "step": 913
    },
    {
      "epoch": 0.69,
      "learning_rate": 4.5688350983358554e-05,
      "loss": 2.0426,
      "step": 914
    },
    {
      "epoch": 0.69,
      "learning_rate": 4.5683623298033286e-05,
      "loss": 1.8825,
      "step": 915
    },
    {
      "epoch": 0.69,
      "learning_rate": 4.5678895612708025e-05,
      "loss": 2.0573,
      "step": 916
    },
    {
      "epoch": 0.69,
      "learning_rate": 4.567416792738276e-05,
      "loss": 2.145,
      "step": 917
    },
    {
      "epoch": 0.69,
      "learning_rate": 4.5669440242057496e-05,
      "loss": 1.8955,
      "step": 918
    },
    {
      "epoch": 0.7,
      "learning_rate": 4.566471255673223e-05,
      "loss": 2.0371,
      "step": 919
    },
    {
      "epoch": 0.7,
      "learning_rate": 4.565998487140696e-05,
      "loss": 1.9649,
      "step": 920
    },
    {
      "epoch": 0.7,
      "learning_rate": 4.56552571860817e-05,
      "loss": 2.1807,
      "step": 921
    },
    {
      "epoch": 0.7,
      "learning_rate": 4.565052950075643e-05,
      "loss": 2.2127,
      "step": 922
    },
    {
      "epoch": 0.7,
      "learning_rate": 4.564580181543116e-05,
      "loss": 2.0379,
      "step": 923
    },
    {
      "epoch": 0.7,
      "learning_rate": 4.56410741301059e-05,
      "loss": 1.9888,
      "step": 924
    },
    {
      "epoch": 0.7,
      "learning_rate": 4.5636346444780634e-05,
      "loss": 2.1463,
      "step": 925
    },
    {
      "epoch": 0.7,
      "learning_rate": 4.563161875945537e-05,
      "loss": 1.9993,
      "step": 926
    },
    {
      "epoch": 0.7,
      "learning_rate": 4.5626891074130105e-05,
      "loss": 2.1424,
      "step": 927
    },
    {
      "epoch": 0.7,
      "learning_rate": 4.5622163388804844e-05,
      "loss": 2.0453,
      "step": 928
    },
    {
      "epoch": 0.7,
      "learning_rate": 4.5617435703479576e-05,
      "loss": 2.0636,
      "step": 929
    },
    {
      "epoch": 0.7,
      "learning_rate": 4.5612708018154315e-05,
      "loss": 2.0491,
      "step": 930
    },
    {
      "epoch": 0.7,
      "learning_rate": 4.560798033282905e-05,
      "loss": 2.0014,
      "step": 931
    },
    {
      "epoch": 0.7,
      "learning_rate": 4.5603252647503785e-05,
      "loss": 2.1246,
      "step": 932
    },
    {
      "epoch": 0.71,
      "learning_rate": 4.559852496217852e-05,
      "loss": 1.9253,
      "step": 933
    },
    {
      "epoch": 0.71,
      "learning_rate": 4.5593797276853256e-05,
      "loss": 2.0153,
      "step": 934
    },
    {
      "epoch": 0.71,
      "learning_rate": 4.558906959152799e-05,
      "loss": 1.9964,
      "step": 935
    },
    {
      "epoch": 0.71,
      "learning_rate": 4.558434190620273e-05,
      "loss": 2.0344,
      "step": 936
    },
    {
      "epoch": 0.71,
      "learning_rate": 4.557961422087746e-05,
      "loss": 2.0204,
      "step": 937
    },
    {
      "epoch": 0.71,
      "learning_rate": 4.55748865355522e-05,
      "loss": 1.9611,
      "step": 938
    },
    {
      "epoch": 0.71,
      "learning_rate": 4.557015885022693e-05,
      "loss": 2.0793,
      "step": 939
    },
    {
      "epoch": 0.71,
      "learning_rate": 4.556543116490167e-05,
      "loss": 2.0296,
      "step": 940
    },
    {
      "epoch": 0.71,
      "learning_rate": 4.55607034795764e-05,
      "loss": 2.0578,
      "step": 941
    },
    {
      "epoch": 0.71,
      "learning_rate": 4.555597579425114e-05,
      "loss": 1.9277,
      "step": 942
    },
    {
      "epoch": 0.71,
      "learning_rate": 4.555124810892587e-05,
      "loss": 2.127,
      "step": 943
    },
    {
      "epoch": 0.71,
      "learning_rate": 4.554652042360061e-05,
      "loss": 1.9804,
      "step": 944
    },
    {
      "epoch": 0.71,
      "learning_rate": 4.554179273827534e-05,
      "loss": 2.001,
      "step": 945
    },
    {
      "epoch": 0.72,
      "learning_rate": 4.553706505295008e-05,
      "loss": 2.117,
      "step": 946
    },
    {
      "epoch": 0.72,
      "learning_rate": 4.5532337367624814e-05,
      "loss": 2.1525,
      "step": 947
    },
    {
      "epoch": 0.72,
      "learning_rate": 4.5527609682299546e-05,
      "loss": 1.8391,
      "step": 948
    },
    {
      "epoch": 0.72,
      "learning_rate": 4.5522881996974285e-05,
      "loss": 1.9776,
      "step": 949
    },
    {
      "epoch": 0.72,
      "learning_rate": 4.551815431164902e-05,
      "loss": 1.9167,
      "step": 950
    },
    {
      "epoch": 0.72,
      "learning_rate": 4.551342662632375e-05,
      "loss": 2.0254,
      "step": 951
    },
    {
      "epoch": 0.72,
      "learning_rate": 4.550869894099849e-05,
      "loss": 2.0757,
      "step": 952
    },
    {
      "epoch": 0.72,
      "learning_rate": 4.550397125567322e-05,
      "loss": 1.9912,
      "step": 953
    },
    {
      "epoch": 0.72,
      "learning_rate": 4.549924357034796e-05,
      "loss": 1.9983,
      "step": 954
    },
    {
      "epoch": 0.72,
      "learning_rate": 4.549451588502269e-05,
      "loss": 1.9235,
      "step": 955
    },
    {
      "epoch": 0.72,
      "learning_rate": 4.548978819969743e-05,
      "loss": 1.8495,
      "step": 956
    },
    {
      "epoch": 0.72,
      "learning_rate": 4.548506051437216e-05,
      "loss": 2.2351,
      "step": 957
    },
    {
      "epoch": 0.72,
      "learning_rate": 4.54803328290469e-05,
      "loss": 1.9736,
      "step": 958
    },
    {
      "epoch": 0.73,
      "learning_rate": 4.547560514372163e-05,
      "loss": 1.8562,
      "step": 959
    },
    {
      "epoch": 0.73,
      "learning_rate": 4.547087745839637e-05,
      "loss": 1.8602,
      "step": 960
    },
    {
      "epoch": 0.73,
      "learning_rate": 4.54661497730711e-05,
      "loss": 2.0561,
      "step": 961
    },
    {
      "epoch": 0.73,
      "learning_rate": 4.546142208774584e-05,
      "loss": 2.0141,
      "step": 962
    },
    {
      "epoch": 0.73,
      "learning_rate": 4.5456694402420574e-05,
      "loss": 2.0357,
      "step": 963
    },
    {
      "epoch": 0.73,
      "learning_rate": 4.545196671709531e-05,
      "loss": 1.9976,
      "step": 964
    },
    {
      "epoch": 0.73,
      "learning_rate": 4.544723903177005e-05,
      "loss": 1.9946,
      "step": 965
    },
    {
      "epoch": 0.73,
      "learning_rate": 4.5442511346444784e-05,
      "loss": 1.9592,
      "step": 966
    },
    {
      "epoch": 0.73,
      "learning_rate": 4.543778366111952e-05,
      "loss": 1.9293,
      "step": 967
    },
    {
      "epoch": 0.73,
      "learning_rate": 4.5433055975794255e-05,
      "loss": 2.101,
      "step": 968
    },
    {
      "epoch": 0.73,
      "learning_rate": 4.5428328290468994e-05,
      "loss": 1.9383,
      "step": 969
    },
    {
      "epoch": 0.73,
      "learning_rate": 4.5423600605143726e-05,
      "loss": 2.0267,
      "step": 970
    },
    {
      "epoch": 0.73,
      "learning_rate": 4.5418872919818465e-05,
      "loss": 2.1624,
      "step": 971
    },
    {
      "epoch": 0.74,
      "learning_rate": 4.54141452344932e-05,
      "loss": 2.1052,
      "step": 972
    },
    {
      "epoch": 0.74,
      "learning_rate": 4.540941754916793e-05,
      "loss": 1.9751,
      "step": 973
    },
    {
      "epoch": 0.74,
      "learning_rate": 4.540468986384267e-05,
      "loss": 2.0394,
      "step": 974
    },
    {
      "epoch": 0.74,
      "learning_rate": 4.53999621785174e-05,
      "loss": 1.9521,
      "step": 975
    },
    {
      "epoch": 0.74,
      "learning_rate": 4.539523449319213e-05,
      "loss": 1.8484,
      "step": 976
    },
    {
      "epoch": 0.74,
      "learning_rate": 4.539050680786687e-05,
      "loss": 2.0445,
      "step": 977
    },
    {
      "epoch": 0.74,
      "learning_rate": 4.53857791225416e-05,
      "loss": 2.0686,
      "step": 978
    },
    {
      "epoch": 0.74,
      "learning_rate": 4.538105143721634e-05,
      "loss": 1.9952,
      "step": 979
    },
    {
      "epoch": 0.74,
      "learning_rate": 4.5376323751891073e-05,
      "loss": 1.973,
      "step": 980
    },
    {
      "epoch": 0.74,
      "learning_rate": 4.537159606656581e-05,
      "loss": 1.8808,
      "step": 981
    },
    {
      "epoch": 0.74,
      "learning_rate": 4.5366868381240544e-05,
      "loss": 1.9153,
      "step": 982
    },
    {
      "epoch": 0.74,
      "learning_rate": 4.536214069591528e-05,
      "loss": 2.0304,
      "step": 983
    },
    {
      "epoch": 0.74,
      "learning_rate": 4.5357413010590015e-05,
      "loss": 1.9612,
      "step": 984
    },
    {
      "epoch": 0.74,
      "learning_rate": 4.5352685325264754e-05,
      "loss": 1.9308,
      "step": 985
    },
    {
      "epoch": 0.75,
      "learning_rate": 4.5347957639939486e-05,
      "loss": 2.1722,
      "step": 986
    },
    {
      "epoch": 0.75,
      "learning_rate": 4.5343229954614225e-05,
      "loss": 1.9808,
      "step": 987
    },
    {
      "epoch": 0.75,
      "learning_rate": 4.533850226928896e-05,
      "loss": 1.978,
      "step": 988
    },
    {
      "epoch": 0.75,
      "learning_rate": 4.5333774583963696e-05,
      "loss": 2.0786,
      "step": 989
    },
    {
      "epoch": 0.75,
      "learning_rate": 4.532904689863843e-05,
      "loss": 1.9463,
      "step": 990
    },
    {
      "epoch": 0.75,
      "learning_rate": 4.532431921331317e-05,
      "loss": 1.8668,
      "step": 991
    },
    {
      "epoch": 0.75,
      "learning_rate": 4.53195915279879e-05,
      "loss": 1.9599,
      "step": 992
    },
    {
      "epoch": 0.75,
      "learning_rate": 4.531486384266264e-05,
      "loss": 1.9149,
      "step": 993
    },
    {
      "epoch": 0.75,
      "learning_rate": 4.531013615733737e-05,
      "loss": 2.0508,
      "step": 994
    },
    {
      "epoch": 0.75,
      "learning_rate": 4.530540847201211e-05,
      "loss": 2.0407,
      "step": 995
    },
    {
      "epoch": 0.75,
      "learning_rate": 4.530068078668684e-05,
      "loss": 1.9946,
      "step": 996
    },
    {
      "epoch": 0.75,
      "learning_rate": 4.529595310136158e-05,
      "loss": 2.0989,
      "step": 997
    },
    {
      "epoch": 0.75,
      "learning_rate": 4.529122541603631e-05,
      "loss": 2.0359,
      "step": 998
    },
    {
      "epoch": 0.76,
      "learning_rate": 4.528649773071105e-05,
      "loss": 2.1105,
      "step": 999
    },
    {
      "epoch": 0.76,
      "learning_rate": 4.528177004538578e-05,
      "loss": 2.059,
      "step": 1000
    },
    {
      "epoch": 0.76,
      "learning_rate": 4.5277042360060514e-05,
      "loss": 2.1039,
      "step": 1001
    },
    {
      "epoch": 0.76,
      "learning_rate": 4.527231467473525e-05,
      "loss": 2.0329,
      "step": 1002
    },
    {
      "epoch": 0.76,
      "learning_rate": 4.5267586989409985e-05,
      "loss": 2.0521,
      "step": 1003
    },
    {
      "epoch": 0.76,
      "learning_rate": 4.526285930408472e-05,
      "loss": 2.0762,
      "step": 1004
    },
    {
      "epoch": 0.76,
      "learning_rate": 4.5258131618759456e-05,
      "loss": 1.9445,
      "step": 1005
    },
    {
      "epoch": 0.76,
      "learning_rate": 4.525340393343419e-05,
      "loss": 1.9423,
      "step": 1006
    },
    {
      "epoch": 0.76,
      "learning_rate": 4.524867624810893e-05,
      "loss": 1.9304,
      "step": 1007
    },
    {
      "epoch": 0.76,
      "learning_rate": 4.524394856278366e-05,
      "loss": 2.1053,
      "step": 1008
    },
    {
      "epoch": 0.76,
      "learning_rate": 4.52392208774584e-05,
      "loss": 2.0303,
      "step": 1009
    },
    {
      "epoch": 0.76,
      "learning_rate": 4.523449319213313e-05,
      "loss": 1.8966,
      "step": 1010
    },
    {
      "epoch": 0.76,
      "learning_rate": 4.522976550680787e-05,
      "loss": 2.0294,
      "step": 1011
    },
    {
      "epoch": 0.77,
      "learning_rate": 4.52250378214826e-05,
      "loss": 2.042,
      "step": 1012
    },
    {
      "epoch": 0.77,
      "learning_rate": 4.522031013615734e-05,
      "loss": 2.049,
      "step": 1013
    },
    {
      "epoch": 0.77,
      "learning_rate": 4.521558245083207e-05,
      "loss": 1.9808,
      "step": 1014
    },
    {
      "epoch": 0.77,
      "learning_rate": 4.521085476550681e-05,
      "loss": 1.934,
      "step": 1015
    },
    {
      "epoch": 0.77,
      "learning_rate": 4.520612708018154e-05,
      "loss": 1.9934,
      "step": 1016
    },
    {
      "epoch": 0.77,
      "learning_rate": 4.520139939485628e-05,
      "loss": 2.0257,
      "step": 1017
    },
    {
      "epoch": 0.77,
      "learning_rate": 4.5196671709531014e-05,
      "loss": 2.0398,
      "step": 1018
    },
    {
      "epoch": 0.77,
      "learning_rate": 4.519194402420575e-05,
      "loss": 2.0448,
      "step": 1019
    },
    {
      "epoch": 0.77,
      "learning_rate": 4.5187216338880485e-05,
      "loss": 2.0634,
      "step": 1020
    },
    {
      "epoch": 0.77,
      "learning_rate": 4.518248865355522e-05,
      "loss": 1.9742,
      "step": 1021
    },
    {
      "epoch": 0.77,
      "learning_rate": 4.5177760968229955e-05,
      "loss": 2.0328,
      "step": 1022
    },
    {
      "epoch": 0.77,
      "learning_rate": 4.5173033282904694e-05,
      "loss": 1.906,
      "step": 1023
    },
    {
      "epoch": 0.77,
      "learning_rate": 4.516830559757943e-05,
      "loss": 2.0367,
      "step": 1024
    },
    {
      "epoch": 0.78,
      "learning_rate": 4.5163577912254165e-05,
      "loss": 2.0815,
      "step": 1025
    },
    {
      "epoch": 0.78,
      "learning_rate": 4.51588502269289e-05,
      "loss": 1.9919,
      "step": 1026
    },
    {
      "epoch": 0.78,
      "learning_rate": 4.5154122541603636e-05,
      "loss": 1.9553,
      "step": 1027
    },
    {
      "epoch": 0.78,
      "learning_rate": 4.514939485627837e-05,
      "loss": 1.9898,
      "step": 1028
    },
    {
      "epoch": 0.78,
      "learning_rate": 4.51446671709531e-05,
      "loss": 1.907,
      "step": 1029
    },
    {
      "epoch": 0.78,
      "learning_rate": 4.513993948562784e-05,
      "loss": 1.9774,
      "step": 1030
    },
    {
      "epoch": 0.78,
      "learning_rate": 4.513521180030257e-05,
      "loss": 1.9224,
      "step": 1031
    },
    {
      "epoch": 0.78,
      "learning_rate": 4.513048411497731e-05,
      "loss": 2.0185,
      "step": 1032
    },
    {
      "epoch": 0.78,
      "learning_rate": 4.512575642965204e-05,
      "loss": 2.099,
      "step": 1033
    },
    {
      "epoch": 0.78,
      "learning_rate": 4.512102874432678e-05,
      "loss": 1.9975,
      "step": 1034
    },
    {
      "epoch": 0.78,
      "learning_rate": 4.511630105900151e-05,
      "loss": 2.0393,
      "step": 1035
    },
    {
      "epoch": 0.78,
      "learning_rate": 4.511157337367625e-05,
      "loss": 2.0027,
      "step": 1036
    },
    {
      "epoch": 0.78,
      "learning_rate": 4.5106845688350984e-05,
      "loss": 1.9128,
      "step": 1037
    },
    {
      "epoch": 0.79,
      "learning_rate": 4.510211800302572e-05,
      "loss": 1.947,
      "step": 1038
    },
    {
      "epoch": 0.79,
      "learning_rate": 4.5097390317700455e-05,
      "loss": 1.9764,
      "step": 1039
    },
    {
      "epoch": 0.79,
      "learning_rate": 4.5092662632375193e-05,
      "loss": 1.934,
      "step": 1040
    },
    {
      "epoch": 0.79,
      "learning_rate": 4.5087934947049926e-05,
      "loss": 1.9708,
      "step": 1041
    },
    {
      "epoch": 0.79,
      "learning_rate": 4.5083207261724664e-05,
      "loss": 2.0312,
      "step": 1042
    },
    {
      "epoch": 0.79,
      "learning_rate": 4.5078479576399396e-05,
      "loss": 1.8832,
      "step": 1043
    },
    {
      "epoch": 0.79,
      "learning_rate": 4.5073751891074135e-05,
      "loss": 2.0414,
      "step": 1044
    },
    {
      "epoch": 0.79,
      "learning_rate": 4.506902420574887e-05,
      "loss": 1.9327,
      "step": 1045
    },
    {
      "epoch": 0.79,
      "learning_rate": 4.5064296520423606e-05,
      "loss": 2.1718,
      "step": 1046
    },
    {
      "epoch": 0.79,
      "learning_rate": 4.505956883509834e-05,
      "loss": 2.0347,
      "step": 1047
    },
    {
      "epoch": 0.79,
      "learning_rate": 4.505484114977308e-05,
      "loss": 2.1503,
      "step": 1048
    },
    {
      "epoch": 0.79,
      "learning_rate": 4.505011346444781e-05,
      "loss": 1.9763,
      "step": 1049
    },
    {
      "epoch": 0.79,
      "learning_rate": 4.504538577912255e-05,
      "loss": 1.9731,
      "step": 1050
    },
    {
      "epoch": 0.79,
      "learning_rate": 4.504065809379728e-05,
      "loss": 2.0236,
      "step": 1051
    },
    {
      "epoch": 0.8,
      "learning_rate": 4.503593040847202e-05,
      "loss": 1.9996,
      "step": 1052
    },
    {
      "epoch": 0.8,
      "learning_rate": 4.503120272314675e-05,
      "loss": 1.8836,
      "step": 1053
    },
    {
      "epoch": 0.8,
      "learning_rate": 4.502647503782148e-05,
      "loss": 2.0622,
      "step": 1054
    },
    {
      "epoch": 0.8,
      "learning_rate": 4.502174735249622e-05,
      "loss": 1.8668,
      "step": 1055
    },
    {
      "epoch": 0.8,
      "learning_rate": 4.5017019667170954e-05,
      "loss": 2.0067,
      "step": 1056
    },
    {
      "epoch": 0.8,
      "learning_rate": 4.5012291981845686e-05,
      "loss": 2.0207,
      "step": 1057
    },
    {
      "epoch": 0.8,
      "learning_rate": 4.5007564296520425e-05,
      "loss": 1.9659,
      "step": 1058
    },
    {
      "epoch": 0.8,
      "learning_rate": 4.500283661119516e-05,
      "loss": 1.8786,
      "step": 1059
    },
    {
      "epoch": 0.8,
      "learning_rate": 4.4998108925869896e-05,
      "loss": 1.9685,
      "step": 1060
    },
    {
      "epoch": 0.8,
      "learning_rate": 4.499338124054463e-05,
      "loss": 1.9907,
      "step": 1061
    },
    {
      "epoch": 0.8,
      "learning_rate": 4.4988653555219367e-05,
      "loss": 1.9875,
      "step": 1062
    },
    {
      "epoch": 0.8,
      "learning_rate": 4.49839258698941e-05,
      "loss": 2.1072,
      "step": 1063
    },
    {
      "epoch": 0.8,
      "learning_rate": 4.497919818456884e-05,
      "loss": 1.9653,
      "step": 1064
    },
    {
      "epoch": 0.81,
      "learning_rate": 4.497447049924357e-05,
      "loss": 2.0709,
      "step": 1065
    },
    {
      "epoch": 0.81,
      "learning_rate": 4.496974281391831e-05,
      "loss": 2.1283,
      "step": 1066
    },
    {
      "epoch": 0.81,
      "learning_rate": 4.496501512859304e-05,
      "loss": 1.9562,
      "step": 1067
    },
    {
      "epoch": 0.81,
      "learning_rate": 4.496028744326778e-05,
      "loss": 1.8642,
      "step": 1068
    },
    {
      "epoch": 0.81,
      "learning_rate": 4.495555975794251e-05,
      "loss": 2.008,
      "step": 1069
    },
    {
      "epoch": 0.81,
      "learning_rate": 4.495083207261725e-05,
      "loss": 2.0039,
      "step": 1070
    },
    {
      "epoch": 0.81,
      "learning_rate": 4.494610438729198e-05,
      "loss": 1.998,
      "step": 1071
    },
    {
      "epoch": 0.81,
      "learning_rate": 4.494137670196672e-05,
      "loss": 1.9448,
      "step": 1072
    },
    {
      "epoch": 0.81,
      "learning_rate": 4.493664901664145e-05,
      "loss": 2.0091,
      "step": 1073
    },
    {
      "epoch": 0.81,
      "learning_rate": 4.493192133131619e-05,
      "loss": 1.9875,
      "step": 1074
    },
    {
      "epoch": 0.81,
      "learning_rate": 4.4927193645990924e-05,
      "loss": 1.8782,
      "step": 1075
    },
    {
      "epoch": 0.81,
      "learning_rate": 4.492246596066566e-05,
      "loss": 1.9391,
      "step": 1076
    },
    {
      "epoch": 0.81,
      "learning_rate": 4.4917738275340395e-05,
      "loss": 1.9873,
      "step": 1077
    },
    {
      "epoch": 0.82,
      "learning_rate": 4.4913010590015134e-05,
      "loss": 2.0604,
      "step": 1078
    },
    {
      "epoch": 0.82,
      "learning_rate": 4.4908282904689866e-05,
      "loss": 1.9803,
      "step": 1079
    },
    {
      "epoch": 0.82,
      "learning_rate": 4.4903555219364605e-05,
      "loss": 2.0415,
      "step": 1080
    },
    {
      "epoch": 0.82,
      "learning_rate": 4.489882753403934e-05,
      "loss": 2.0557,
      "step": 1081
    },
    {
      "epoch": 0.82,
      "learning_rate": 4.489409984871407e-05,
      "loss": 2.1134,
      "step": 1082
    },
    {
      "epoch": 0.82,
      "learning_rate": 4.488937216338881e-05,
      "loss": 1.9881,
      "step": 1083
    },
    {
      "epoch": 0.82,
      "learning_rate": 4.488464447806354e-05,
      "loss": 1.9538,
      "step": 1084
    },
    {
      "epoch": 0.82,
      "learning_rate": 4.487991679273828e-05,
      "loss": 1.8229,
      "step": 1085
    },
    {
      "epoch": 0.82,
      "learning_rate": 4.487518910741301e-05,
      "loss": 2.0652,
      "step": 1086
    },
    {
      "epoch": 0.82,
      "learning_rate": 4.487046142208775e-05,
      "loss": 2.014,
      "step": 1087
    },
    {
      "epoch": 0.82,
      "learning_rate": 4.486573373676248e-05,
      "loss": 2.0786,
      "step": 1088
    },
    {
      "epoch": 0.82,
      "learning_rate": 4.486100605143722e-05,
      "loss": 2.0479,
      "step": 1089
    },
    {
      "epoch": 0.82,
      "learning_rate": 4.485627836611195e-05,
      "loss": 1.8588,
      "step": 1090
    },
    {
      "epoch": 0.83,
      "learning_rate": 4.485155068078669e-05,
      "loss": 2.0716,
      "step": 1091
    },
    {
      "epoch": 0.83,
      "learning_rate": 4.484682299546142e-05,
      "loss": 2.0001,
      "step": 1092
    },
    {
      "epoch": 0.83,
      "learning_rate": 4.484209531013616e-05,
      "loss": 2.0745,
      "step": 1093
    },
    {
      "epoch": 0.83,
      "learning_rate": 4.4837367624810894e-05,
      "loss": 1.9944,
      "step": 1094
    },
    {
      "epoch": 0.83,
      "learning_rate": 4.483263993948563e-05,
      "loss": 1.9736,
      "step": 1095
    },
    {
      "epoch": 0.83,
      "learning_rate": 4.4827912254160365e-05,
      "loss": 2.0684,
      "step": 1096
    },
    {
      "epoch": 0.83,
      "learning_rate": 4.4823184568835104e-05,
      "loss": 2.0382,
      "step": 1097
    },
    {
      "epoch": 0.83,
      "learning_rate": 4.4818456883509836e-05,
      "loss": 1.9479,
      "step": 1098
    },
    {
      "epoch": 0.83,
      "learning_rate": 4.4813729198184575e-05,
      "loss": 2.1523,
      "step": 1099
    },
    {
      "epoch": 0.83,
      "learning_rate": 4.480900151285931e-05,
      "loss": 1.9328,
      "step": 1100
    },
    {
      "epoch": 0.83,
      "learning_rate": 4.4804273827534046e-05,
      "loss": 2.0274,
      "step": 1101
    },
    {
      "epoch": 0.83,
      "learning_rate": 4.479954614220878e-05,
      "loss": 2.0468,
      "step": 1102
    },
    {
      "epoch": 0.83,
      "learning_rate": 4.4794818456883517e-05,
      "loss": 2.0977,
      "step": 1103
    },
    {
      "epoch": 0.83,
      "learning_rate": 4.479009077155825e-05,
      "loss": 1.9469,
      "step": 1104
    },
    {
      "epoch": 0.84,
      "learning_rate": 4.478536308623299e-05,
      "loss": 1.9386,
      "step": 1105
    },
    {
      "epoch": 0.84,
      "learning_rate": 4.478063540090772e-05,
      "loss": 1.9917,
      "step": 1106
    },
    {
      "epoch": 0.84,
      "learning_rate": 4.477590771558245e-05,
      "loss": 2.1049,
      "step": 1107
    },
    {
      "epoch": 0.84,
      "learning_rate": 4.477118003025719e-05,
      "loss": 1.9923,
      "step": 1108
    },
    {
      "epoch": 0.84,
      "learning_rate": 4.476645234493192e-05,
      "loss": 2.0673,
      "step": 1109
    },
    {
      "epoch": 0.84,
      "learning_rate": 4.4761724659606654e-05,
      "loss": 1.9602,
      "step": 1110
    },
    {
      "epoch": 0.84,
      "learning_rate": 4.475699697428139e-05,
      "loss": 1.9975,
      "step": 1111
    },
    {
      "epoch": 0.84,
      "learning_rate": 4.4752269288956125e-05,
      "loss": 1.9682,
      "step": 1112
    },
    {
      "epoch": 0.84,
      "learning_rate": 4.4747541603630864e-05,
      "loss": 2.0077,
      "step": 1113
    },
    {
      "epoch": 0.84,
      "learning_rate": 4.4742813918305596e-05,
      "loss": 2.1605,
      "step": 1114
    },
    {
      "epoch": 0.84,
      "learning_rate": 4.4738086232980335e-05,
      "loss": 2.0586,
      "step": 1115
    },
    {
      "epoch": 0.84,
      "learning_rate": 4.473335854765507e-05,
      "loss": 2.1005,
      "step": 1116
    },
    {
      "epoch": 0.84,
      "learning_rate": 4.4728630862329806e-05,
      "loss": 2.2024,
      "step": 1117
    },
    {
      "epoch": 0.85,
      "learning_rate": 4.472390317700454e-05,
      "loss": 2.0624,
      "step": 1118
    },
    {
      "epoch": 0.85,
      "learning_rate": 4.471917549167928e-05,
      "loss": 1.9187,
      "step": 1119
    },
    {
      "epoch": 0.85,
      "learning_rate": 4.471444780635401e-05,
      "loss": 2.1008,
      "step": 1120
    },
    {
      "epoch": 0.85,
      "learning_rate": 4.470972012102875e-05,
      "loss": 2.2395,
      "step": 1121
    },
    {
      "epoch": 0.85,
      "learning_rate": 4.470499243570348e-05,
      "loss": 2.0519,
      "step": 1122
    },
    {
      "epoch": 0.85,
      "learning_rate": 4.470026475037822e-05,
      "loss": 1.9914,
      "step": 1123
    },
    {
      "epoch": 0.85,
      "learning_rate": 4.469553706505295e-05,
      "loss": 2.0299,
      "step": 1124
    },
    {
      "epoch": 0.85,
      "learning_rate": 4.469080937972769e-05,
      "loss": 1.921,
      "step": 1125
    },
    {
      "epoch": 0.85,
      "learning_rate": 4.468608169440242e-05,
      "loss": 1.9605,
      "step": 1126
    },
    {
      "epoch": 0.85,
      "learning_rate": 4.468135400907716e-05,
      "loss": 1.9685,
      "step": 1127
    },
    {
      "epoch": 0.85,
      "learning_rate": 4.467662632375189e-05,
      "loss": 2.0456,
      "step": 1128
    },
    {
      "epoch": 0.85,
      "learning_rate": 4.467189863842663e-05,
      "loss": 1.8312,
      "step": 1129
    },
    {
      "epoch": 0.85,
      "learning_rate": 4.4667170953101363e-05,
      "loss": 2.1349,
      "step": 1130
    },
    {
      "epoch": 0.86,
      "learning_rate": 4.46624432677761e-05,
      "loss": 2.1713,
      "step": 1131
    },
    {
      "epoch": 0.86,
      "learning_rate": 4.4657715582450834e-05,
      "loss": 2.0393,
      "step": 1132
    },
    {
      "epoch": 0.86,
      "learning_rate": 4.465298789712557e-05,
      "loss": 2.1671,
      "step": 1133
    },
    {
      "epoch": 0.86,
      "learning_rate": 4.4648260211800305e-05,
      "loss": 1.9418,
      "step": 1134
    },
    {
      "epoch": 0.86,
      "learning_rate": 4.464353252647504e-05,
      "loss": 1.8798,
      "step": 1135
    },
    {
      "epoch": 0.86,
      "learning_rate": 4.4638804841149776e-05,
      "loss": 2.025,
      "step": 1136
    },
    {
      "epoch": 0.86,
      "learning_rate": 4.463407715582451e-05,
      "loss": 1.9963,
      "step": 1137
    },
    {
      "epoch": 0.86,
      "learning_rate": 4.462934947049924e-05,
      "loss": 1.9496,
      "step": 1138
    },
    {
      "epoch": 0.86,
      "learning_rate": 4.462462178517398e-05,
      "loss": 1.8681,
      "step": 1139
    },
    {
      "epoch": 0.86,
      "learning_rate": 4.461989409984871e-05,
      "loss": 1.9093,
      "step": 1140
    },
    {
      "epoch": 0.86,
      "learning_rate": 4.461516641452345e-05,
      "loss": 1.9739,
      "step": 1141
    },
    {
      "epoch": 0.86,
      "learning_rate": 4.461043872919818e-05,
      "loss": 1.9765,
      "step": 1142
    },
    {
      "epoch": 0.86,
      "learning_rate": 4.460571104387292e-05,
      "loss": 2.0152,
      "step": 1143
    },
    {
      "epoch": 0.87,
      "learning_rate": 4.460098335854766e-05,
      "loss": 1.9865,
      "step": 1144
    },
    {
      "epoch": 0.87,
      "learning_rate": 4.459625567322239e-05,
      "loss": 1.8557,
      "step": 1145
    },
    {
      "epoch": 0.87,
      "learning_rate": 4.459152798789713e-05,
      "loss": 2.0907,
      "step": 1146
    },
    {
      "epoch": 0.87,
      "learning_rate": 4.458680030257186e-05,
      "loss": 1.9935,
      "step": 1147
    },
    {
      "epoch": 0.87,
      "learning_rate": 4.45820726172466e-05,
      "loss": 1.9475,
      "step": 1148
    },
    {
      "epoch": 0.87,
      "learning_rate": 4.4577344931921334e-05,
      "loss": 2.0259,
      "step": 1149
    },
    {
      "epoch": 0.87,
      "learning_rate": 4.457261724659607e-05,
      "loss": 1.8934,
      "step": 1150
    },
    {
      "epoch": 0.87,
      "learning_rate": 4.4567889561270804e-05,
      "loss": 2.0507,
      "step": 1151
    },
    {
      "epoch": 0.87,
      "learning_rate": 4.456316187594554e-05,
      "loss": 1.9941,
      "step": 1152
    },
    {
      "epoch": 0.87,
      "learning_rate": 4.4558434190620275e-05,
      "loss": 1.9518,
      "step": 1153
    },
    {
      "epoch": 0.87,
      "learning_rate": 4.4553706505295014e-05,
      "loss": 2.0914,
      "step": 1154
    },
    {
      "epoch": 0.87,
      "learning_rate": 4.4548978819969746e-05,
      "loss": 1.9503,
      "step": 1155
    },
    {
      "epoch": 0.87,
      "learning_rate": 4.4544251134644485e-05,
      "loss": 2.0164,
      "step": 1156
    },
    {
      "epoch": 0.88,
      "learning_rate": 4.453952344931922e-05,
      "loss": 2.0177,
      "step": 1157
    },
    {
      "epoch": 0.88,
      "learning_rate": 4.4534795763993956e-05,
      "loss": 2.0289,
      "step": 1158
    },
    {
      "epoch": 0.88,
      "learning_rate": 4.453006807866869e-05,
      "loss": 1.9981,
      "step": 1159
    },
    {
      "epoch": 0.88,
      "learning_rate": 4.452534039334342e-05,
      "loss": 2.1372,
      "step": 1160
    },
    {
      "epoch": 0.88,
      "learning_rate": 4.452061270801816e-05,
      "loss": 2.0588,
      "step": 1161
    },
    {
      "epoch": 0.88,
      "learning_rate": 4.451588502269289e-05,
      "loss": 1.9849,
      "step": 1162
    },
    {
      "epoch": 0.88,
      "learning_rate": 4.451115733736762e-05,
      "loss": 1.9594,
      "step": 1163
    },
    {
      "epoch": 0.88,
      "learning_rate": 4.450642965204236e-05,
      "loss": 1.9895,
      "step": 1164
    },
    {
      "epoch": 0.88,
      "learning_rate": 4.4501701966717094e-05,
      "loss": 2.0632,
      "step": 1165
    },
    {
      "epoch": 0.88,
      "learning_rate": 4.449697428139183e-05,
      "loss": 1.9899,
      "step": 1166
    },
    {
      "epoch": 0.88,
      "learning_rate": 4.4492246596066565e-05,
      "loss": 1.9934,
      "step": 1167
    },
    {
      "epoch": 0.88,
      "learning_rate": 4.4487518910741304e-05,
      "loss": 2.0015,
      "step": 1168
    },
    {
      "epoch": 0.88,
      "learning_rate": 4.4482791225416036e-05,
      "loss": 1.9206,
      "step": 1169
    },
    {
      "epoch": 0.88,
      "learning_rate": 4.4478063540090775e-05,
      "loss": 2.0638,
      "step": 1170
    },
    {
      "epoch": 0.89,
      "learning_rate": 4.4473335854765507e-05,
      "loss": 2.0885,
      "step": 1171
    },
    {
      "epoch": 0.89,
      "learning_rate": 4.4468608169440245e-05,
      "loss": 2.1329,
      "step": 1172
    },
    {
      "epoch": 0.89,
      "learning_rate": 4.446388048411498e-05,
      "loss": 2.0848,
      "step": 1173
    },
    {
      "epoch": 0.89,
      "learning_rate": 4.4459152798789716e-05,
      "loss": 2.0424,
      "step": 1174
    },
    {
      "epoch": 0.89,
      "learning_rate": 4.445442511346445e-05,
      "loss": 2.0171,
      "step": 1175
    },
    {
      "epoch": 0.89,
      "learning_rate": 4.444969742813919e-05,
      "loss": 1.9713,
      "step": 1176
    },
    {
      "epoch": 0.89,
      "learning_rate": 4.444496974281392e-05,
      "loss": 1.9072,
      "step": 1177
    },
    {
      "epoch": 0.89,
      "learning_rate": 4.444024205748866e-05,
      "loss": 2.1101,
      "step": 1178
    },
    {
      "epoch": 0.89,
      "learning_rate": 4.443551437216339e-05,
      "loss": 2.1196,
      "step": 1179
    },
    {
      "epoch": 0.89,
      "learning_rate": 4.443078668683813e-05,
      "loss": 2.0584,
      "step": 1180
    },
    {
      "epoch": 0.89,
      "learning_rate": 4.442605900151286e-05,
      "loss": 2.0529,
      "step": 1181
    },
    {
      "epoch": 0.89,
      "learning_rate": 4.44213313161876e-05,
      "loss": 2.1246,
      "step": 1182
    },
    {
      "epoch": 0.89,
      "learning_rate": 4.441660363086233e-05,
      "loss": 1.9808,
      "step": 1183
    },
    {
      "epoch": 0.9,
      "learning_rate": 4.441187594553707e-05,
      "loss": 2.0232,
      "step": 1184
    },
    {
      "epoch": 0.9,
      "learning_rate": 4.44071482602118e-05,
      "loss": 2.035,
      "step": 1185
    },
    {
      "epoch": 0.9,
      "learning_rate": 4.440242057488654e-05,
      "loss": 1.9754,
      "step": 1186
    },
    {
      "epoch": 0.9,
      "learning_rate": 4.4397692889561274e-05,
      "loss": 1.9198,
      "step": 1187
    },
    {
      "epoch": 0.9,
      "learning_rate": 4.4392965204236006e-05,
      "loss": 2.0763,
      "step": 1188
    },
    {
      "epoch": 0.9,
      "learning_rate": 4.4388237518910745e-05,
      "loss": 1.9196,
      "step": 1189
    },
    {
      "epoch": 0.9,
      "learning_rate": 4.438350983358548e-05,
      "loss": 1.9901,
      "step": 1190
    },
    {
      "epoch": 0.9,
      "learning_rate": 4.437878214826021e-05,
      "loss": 1.9,
      "step": 1191
    },
    {
      "epoch": 0.9,
      "learning_rate": 4.437405446293495e-05,
      "loss": 2.1058,
      "step": 1192
    },
    {
      "epoch": 0.9,
      "learning_rate": 4.436932677760968e-05,
      "loss": 2.0535,
      "step": 1193
    },
    {
      "epoch": 0.9,
      "learning_rate": 4.436459909228442e-05,
      "loss": 1.9386,
      "step": 1194
    },
    {
      "epoch": 0.9,
      "learning_rate": 4.435987140695915e-05,
      "loss": 2.0213,
      "step": 1195
    },
    {
      "epoch": 0.9,
      "learning_rate": 4.435514372163389e-05,
      "loss": 1.8508,
      "step": 1196
    },
    {
      "epoch": 0.91,
      "learning_rate": 4.435041603630862e-05,
      "loss": 1.9433,
      "step": 1197
    },
    {
      "epoch": 0.91,
      "learning_rate": 4.434568835098336e-05,
      "loss": 2.101,
      "step": 1198
    },
    {
      "epoch": 0.91,
      "learning_rate": 4.434096066565809e-05,
      "loss": 2.024,
      "step": 1199
    },
    {
      "epoch": 0.91,
      "learning_rate": 4.433623298033283e-05,
      "loss": 2.0493,
      "step": 1200
    },
    {
      "epoch": 0.91,
      "learning_rate": 4.433150529500756e-05,
      "loss": 1.9834,
      "step": 1201
    },
    {
      "epoch": 0.91,
      "learning_rate": 4.43267776096823e-05,
      "loss": 1.9577,
      "step": 1202
    },
    {
      "epoch": 0.91,
      "learning_rate": 4.4322049924357034e-05,
      "loss": 1.9897,
      "step": 1203
    },
    {
      "epoch": 0.91,
      "learning_rate": 4.431732223903177e-05,
      "loss": 2.0251,
      "step": 1204
    },
    {
      "epoch": 0.91,
      "learning_rate": 4.431259455370651e-05,
      "loss": 1.9344,
      "step": 1205
    },
    {
      "epoch": 0.91,
      "learning_rate": 4.4307866868381244e-05,
      "loss": 2.041,
      "step": 1206
    },
    {
      "epoch": 0.91,
      "learning_rate": 4.430313918305598e-05,
      "loss": 2.0699,
      "step": 1207
    },
    {
      "epoch": 0.91,
      "learning_rate": 4.4298411497730715e-05,
      "loss": 1.8758,
      "step": 1208
    },
    {
      "epoch": 0.91,
      "learning_rate": 4.4293683812405454e-05,
      "loss": 1.9562,
      "step": 1209
    },
    {
      "epoch": 0.92,
      "learning_rate": 4.4288956127080186e-05,
      "loss": 2.1196,
      "step": 1210
    },
    {
      "epoch": 0.92,
      "learning_rate": 4.4284228441754924e-05,
      "loss": 2.0394,
      "step": 1211
    },
    {
      "epoch": 0.92,
      "learning_rate": 4.4279500756429657e-05,
      "loss": 2.0055,
      "step": 1212
    },
    {
      "epoch": 0.92,
      "learning_rate": 4.427477307110439e-05,
      "loss": 2.0228,
      "step": 1213
    },
    {
      "epoch": 0.92,
      "learning_rate": 4.427004538577913e-05,
      "loss": 1.9039,
      "step": 1214
    },
    {
      "epoch": 0.92,
      "learning_rate": 4.426531770045386e-05,
      "loss": 2.0931,
      "step": 1215
    },
    {
      "epoch": 0.92,
      "learning_rate": 4.426059001512859e-05,
      "loss": 1.9317,
      "step": 1216
    },
    {
      "epoch": 0.92,
      "learning_rate": 4.425586232980333e-05,
      "loss": 2.0344,
      "step": 1217
    },
    {
      "epoch": 0.92,
      "learning_rate": 4.425113464447806e-05,
      "loss": 1.9156,
      "step": 1218
    },
    {
      "epoch": 0.92,
      "learning_rate": 4.42464069591528e-05,
      "loss": 1.9891,
      "step": 1219
    },
    {
      "epoch": 0.92,
      "learning_rate": 4.424167927382753e-05,
      "loss": 2.1506,
      "step": 1220
    },
    {
      "epoch": 0.92,
      "learning_rate": 4.423695158850227e-05,
      "loss": 2.144,
      "step": 1221
    },
    {
      "epoch": 0.92,
      "learning_rate": 4.4232223903177004e-05,
      "loss": 2.1246,
      "step": 1222
    },
    {
      "epoch": 0.92,
      "learning_rate": 4.422749621785174e-05,
      "loss": 1.9525,
      "step": 1223
    },
    {
      "epoch": 0.93,
      "learning_rate": 4.4222768532526475e-05,
      "loss": 1.9795,
      "step": 1224
    },
    {
      "epoch": 0.93,
      "learning_rate": 4.4218040847201214e-05,
      "loss": 2.0335,
      "step": 1225
    },
    {
      "epoch": 0.93,
      "learning_rate": 4.4213313161875946e-05,
      "loss": 2.075,
      "step": 1226
    },
    {
      "epoch": 0.93,
      "learning_rate": 4.4208585476550685e-05,
      "loss": 2.0182,
      "step": 1227
    },
    {
      "epoch": 0.93,
      "learning_rate": 4.420385779122542e-05,
      "loss": 2.0052,
      "step": 1228
    },
    {
      "epoch": 0.93,
      "learning_rate": 4.4199130105900156e-05,
      "loss": 1.9375,
      "step": 1229
    },
    {
      "epoch": 0.93,
      "learning_rate": 4.419440242057489e-05,
      "loss": 1.9774,
      "step": 1230
    },
    {
      "epoch": 0.93,
      "learning_rate": 4.418967473524963e-05,
      "loss": 1.9338,
      "step": 1231
    },
    {
      "epoch": 0.93,
      "learning_rate": 4.418494704992436e-05,
      "loss": 2.0199,
      "step": 1232
    },
    {
      "epoch": 0.93,
      "learning_rate": 4.41802193645991e-05,
      "loss": 2.0348,
      "step": 1233
    },
    {
      "epoch": 0.93,
      "learning_rate": 4.417549167927383e-05,
      "loss": 2.0631,
      "step": 1234
    },
    {
      "epoch": 0.93,
      "learning_rate": 4.417076399394857e-05,
      "loss": 1.989,
      "step": 1235
    },
    {
      "epoch": 0.93,
      "learning_rate": 4.41660363086233e-05,
      "loss": 2.03,
      "step": 1236
    },
    {
      "epoch": 0.94,
      "learning_rate": 4.416130862329804e-05,
      "loss": 2.082,
      "step": 1237
    },
    {
      "epoch": 0.94,
      "learning_rate": 4.415658093797277e-05,
      "loss": 1.9478,
      "step": 1238
    },
    {
      "epoch": 0.94,
      "learning_rate": 4.415185325264751e-05,
      "loss": 2.0342,
      "step": 1239
    },
    {
      "epoch": 0.94,
      "learning_rate": 4.414712556732224e-05,
      "loss": 2.031,
      "step": 1240
    },
    {
      "epoch": 0.94,
      "learning_rate": 4.4142397881996974e-05,
      "loss": 2.1617,
      "step": 1241
    },
    {
      "epoch": 0.94,
      "learning_rate": 4.413767019667171e-05,
      "loss": 1.9573,
      "step": 1242
    },
    {
      "epoch": 0.94,
      "learning_rate": 4.4132942511346445e-05,
      "loss": 1.9324,
      "step": 1243
    },
    {
      "epoch": 0.94,
      "learning_rate": 4.412821482602118e-05,
      "loss": 2.0879,
      "step": 1244
    },
    {
      "epoch": 0.94,
      "learning_rate": 4.4123487140695916e-05,
      "loss": 2.064,
      "step": 1245
    },
    {
      "epoch": 0.94,
      "learning_rate": 4.411875945537065e-05,
      "loss": 2.0583,
      "step": 1246
    },
    {
      "epoch": 0.94,
      "learning_rate": 4.411403177004539e-05,
      "loss": 1.9675,
      "step": 1247
    },
    {
      "epoch": 0.94,
      "learning_rate": 4.410930408472012e-05,
      "loss": 2.0356,
      "step": 1248
    },
    {
      "epoch": 0.94,
      "learning_rate": 4.410457639939486e-05,
      "loss": 1.9396,
      "step": 1249
    },
    {
      "epoch": 0.95,
      "learning_rate": 4.409984871406959e-05,
      "loss": 2.121,
      "step": 1250
    },
    {
      "epoch": 0.95,
      "learning_rate": 4.409512102874433e-05,
      "loss": 1.9724,
      "step": 1251
    },
    {
      "epoch": 0.95,
      "learning_rate": 4.409039334341906e-05,
      "loss": 1.9831,
      "step": 1252
    },
    {
      "epoch": 0.95,
      "learning_rate": 4.40856656580938e-05,
      "loss": 1.9637,
      "step": 1253
    },
    {
      "epoch": 0.95,
      "learning_rate": 4.408093797276853e-05,
      "loss": 2.1036,
      "step": 1254
    },
    {
      "epoch": 0.95,
      "learning_rate": 4.407621028744327e-05,
      "loss": 2.0063,
      "step": 1255
    },
    {
      "epoch": 0.95,
      "learning_rate": 4.4071482602118e-05,
      "loss": 2.0036,
      "step": 1256
    },
    {
      "epoch": 0.95,
      "learning_rate": 4.406675491679274e-05,
      "loss": 1.9162,
      "step": 1257
    },
    {
      "epoch": 0.95,
      "learning_rate": 4.4062027231467474e-05,
      "loss": 2.0343,
      "step": 1258
    },
    {
      "epoch": 0.95,
      "learning_rate": 4.405729954614221e-05,
      "loss": 1.955,
      "step": 1259
    },
    {
      "epoch": 0.95,
      "learning_rate": 4.4052571860816944e-05,
      "loss": 1.893,
      "step": 1260
    },
    {
      "epoch": 0.95,
      "learning_rate": 4.404784417549168e-05,
      "loss": 2.0366,
      "step": 1261
    },
    {
      "epoch": 0.95,
      "learning_rate": 4.4043116490166415e-05,
      "loss": 1.8854,
      "step": 1262
    },
    {
      "epoch": 0.96,
      "learning_rate": 4.4038388804841154e-05,
      "loss": 1.9901,
      "step": 1263
    },
    {
      "epoch": 0.96,
      "learning_rate": 4.403366111951589e-05,
      "loss": 1.99,
      "step": 1264
    },
    {
      "epoch": 0.96,
      "learning_rate": 4.4028933434190625e-05,
      "loss": 2.0581,
      "step": 1265
    },
    {
      "epoch": 0.96,
      "learning_rate": 4.402420574886536e-05,
      "loss": 2.0881,
      "step": 1266
    },
    {
      "epoch": 0.96,
      "learning_rate": 4.4019478063540096e-05,
      "loss": 1.965,
      "step": 1267
    },
    {
      "epoch": 0.96,
      "learning_rate": 4.401475037821483e-05,
      "loss": 1.9453,
      "step": 1268
    },
    {
      "epoch": 0.96,
      "learning_rate": 4.401002269288956e-05,
      "loss": 1.8581,
      "step": 1269
    },
    {
      "epoch": 0.96,
      "learning_rate": 4.40052950075643e-05,
      "loss": 2.0035,
      "step": 1270
    },
    {
      "epoch": 0.96,
      "learning_rate": 4.400056732223903e-05,
      "loss": 1.9703,
      "step": 1271
    },
    {
      "epoch": 0.96,
      "learning_rate": 4.399583963691377e-05,
      "loss": 1.9211,
      "step": 1272
    },
    {
      "epoch": 0.96,
      "learning_rate": 4.39911119515885e-05,
      "loss": 2.043,
      "step": 1273
    },
    {
      "epoch": 0.96,
      "learning_rate": 4.398638426626324e-05,
      "loss": 2.0588,
      "step": 1274
    },
    {
      "epoch": 0.96,
      "learning_rate": 4.398165658093797e-05,
      "loss": 1.925,
      "step": 1275
    },
    {
      "epoch": 0.97,
      "learning_rate": 4.397692889561271e-05,
      "loss": 2.014,
      "step": 1276
    },
    {
      "epoch": 0.97,
      "learning_rate": 4.3972201210287444e-05,
      "loss": 2.0509,
      "step": 1277
    },
    {
      "epoch": 0.97,
      "learning_rate": 4.396747352496218e-05,
      "loss": 2.0194,
      "step": 1278
    },
    {
      "epoch": 0.97,
      "learning_rate": 4.3962745839636915e-05,
      "loss": 1.9255,
      "step": 1279
    },
    {
      "epoch": 0.97,
      "learning_rate": 4.3958018154311653e-05,
      "loss": 1.9734,
      "step": 1280
    },
    {
      "epoch": 0.97,
      "learning_rate": 4.3953290468986385e-05,
      "loss": 2.1426,
      "step": 1281
    },
    {
      "epoch": 0.97,
      "learning_rate": 4.3948562783661124e-05,
      "loss": 1.9921,
      "step": 1282
    },
    {
      "epoch": 0.97,
      "learning_rate": 4.3943835098335856e-05,
      "loss": 2.0806,
      "step": 1283
    },
    {
      "epoch": 0.97,
      "learning_rate": 4.3939107413010595e-05,
      "loss": 2.013,
      "step": 1284
    },
    {
      "epoch": 0.97,
      "learning_rate": 4.393437972768533e-05,
      "loss": 1.9267,
      "step": 1285
    },
    {
      "epoch": 0.97,
      "learning_rate": 4.3929652042360066e-05,
      "loss": 2.0773,
      "step": 1286
    },
    {
      "epoch": 0.97,
      "learning_rate": 4.39249243570348e-05,
      "loss": 1.9651,
      "step": 1287
    },
    {
      "epoch": 0.97,
      "learning_rate": 4.392019667170954e-05,
      "loss": 2.0362,
      "step": 1288
    },
    {
      "epoch": 0.97,
      "learning_rate": 4.391546898638427e-05,
      "loss": 2.1191,
      "step": 1289
    },
    {
      "epoch": 0.98,
      "learning_rate": 4.391074130105901e-05,
      "loss": 1.9121,
      "step": 1290
    },
    {
      "epoch": 0.98,
      "learning_rate": 4.390601361573374e-05,
      "loss": 1.9734,
      "step": 1291
    },
    {
      "epoch": 0.98,
      "learning_rate": 4.390128593040848e-05,
      "loss": 2.085,
      "step": 1292
    },
    {
      "epoch": 0.98,
      "learning_rate": 4.389655824508321e-05,
      "loss": 2.0894,
      "step": 1293
    },
    {
      "epoch": 0.98,
      "learning_rate": 4.389183055975794e-05,
      "loss": 1.9982,
      "step": 1294
    },
    {
      "epoch": 0.98,
      "learning_rate": 4.388710287443268e-05,
      "loss": 1.9522,
      "step": 1295
    },
    {
      "epoch": 0.98,
      "learning_rate": 4.3882375189107414e-05,
      "loss": 1.982,
      "step": 1296
    },
    {
      "epoch": 0.98,
      "learning_rate": 4.3877647503782146e-05,
      "loss": 2.0059,
      "step": 1297
    },
    {
      "epoch": 0.98,
      "learning_rate": 4.3872919818456885e-05,
      "loss": 1.9535,
      "step": 1298
    },
    {
      "epoch": 0.98,
      "learning_rate": 4.386819213313162e-05,
      "loss": 2.0226,
      "step": 1299
    },
    {
      "epoch": 0.98,
      "learning_rate": 4.3863464447806356e-05,
      "loss": 1.9611,
      "step": 1300
    },
    {
      "epoch": 0.98,
      "learning_rate": 4.385873676248109e-05,
      "loss": 1.9974,
      "step": 1301
    },
    {
      "epoch": 0.98,
      "learning_rate": 4.3854009077155826e-05,
      "loss": 1.9908,
      "step": 1302
    },
    {
      "epoch": 0.99,
      "learning_rate": 4.384928139183056e-05,
      "loss": 1.9709,
      "step": 1303
    },
    {
      "epoch": 0.99,
      "learning_rate": 4.38445537065053e-05,
      "loss": 1.9916,
      "step": 1304
    },
    {
      "epoch": 0.99,
      "learning_rate": 4.383982602118003e-05,
      "loss": 2.1279,
      "step": 1305
    },
    {
      "epoch": 0.99,
      "learning_rate": 4.383509833585477e-05,
      "loss": 1.9637,
      "step": 1306
    },
    {
      "epoch": 0.99,
      "learning_rate": 4.38303706505295e-05,
      "loss": 2.0203,
      "step": 1307
    },
    {
      "epoch": 0.99,
      "learning_rate": 4.382564296520424e-05,
      "loss": 1.9688,
      "step": 1308
    },
    {
      "epoch": 0.99,
      "learning_rate": 4.382091527987897e-05,
      "loss": 2.0334,
      "step": 1309
    },
    {
      "epoch": 0.99,
      "learning_rate": 4.381618759455371e-05,
      "loss": 1.928,
      "step": 1310
    },
    {
      "epoch": 0.99,
      "learning_rate": 4.381145990922844e-05,
      "loss": 2.0224,
      "step": 1311
    },
    {
      "epoch": 0.99,
      "learning_rate": 4.380673222390318e-05,
      "loss": 1.9813,
      "step": 1312
    },
    {
      "epoch": 0.99,
      "learning_rate": 4.380200453857791e-05,
      "loss": 1.9812,
      "step": 1313
    },
    {
      "epoch": 0.99,
      "learning_rate": 4.379727685325265e-05,
      "loss": 1.8933,
      "step": 1314
    },
    {
      "epoch": 0.99,
      "learning_rate": 4.3792549167927384e-05,
      "loss": 1.9376,
      "step": 1315
    },
    {
      "epoch": 1.0,
      "learning_rate": 4.378782148260212e-05,
      "loss": 2.0272,
      "step": 1316
    },
    {
      "epoch": 1.0,
      "learning_rate": 4.3783093797276855e-05,
      "loss": 1.9023,
      "step": 1317
    },
    {
      "epoch": 1.0,
      "learning_rate": 4.3778366111951594e-05,
      "loss": 2.0011,
      "step": 1318
    },
    {
      "epoch": 1.0,
      "learning_rate": 4.3773638426626326e-05,
      "loss": 1.9494,
      "step": 1319
    },
    {
      "epoch": 1.0,
      "learning_rate": 4.3768910741301065e-05,
      "loss": 1.9097,
      "step": 1320
    },
    {
      "epoch": 1.0,
      "learning_rate": 4.3764183055975797e-05,
      "loss": 2.0886,
      "step": 1321
    },
    {
      "epoch": 1.0,
      "learning_rate": 4.375945537065053e-05,
      "loss": 2.1207,
      "step": 1322
    },
    {
      "epoch": 1.0,
      "learning_rate": 4.375472768532527e-05,
      "loss": 1.3393,
      "step": 1323
    },
    {
      "epoch": 1.0,
      "learning_rate": 4.375e-05,
      "loss": 1.1356,
      "step": 1324
    },
    {
      "epoch": 1.0,
      "learning_rate": 4.374527231467474e-05,
      "loss": 0.9773,
      "step": 1325
    },
    {
      "epoch": 1.0,
      "learning_rate": 4.374054462934947e-05,
      "loss": 0.9422,
      "step": 1326
    },
    {
      "epoch": 1.0,
      "learning_rate": 4.373581694402421e-05,
      "loss": 0.8269,
      "step": 1327
    },
    {
      "epoch": 1.0,
      "learning_rate": 4.373108925869894e-05,
      "loss": 0.7969,
      "step": 1328
    },
    {
      "epoch": 1.01,
      "learning_rate": 4.372636157337368e-05,
      "loss": 0.9238,
      "step": 1329
    },
    {
      "epoch": 1.01,
      "learning_rate": 4.372163388804841e-05,
      "loss": 0.953,
      "step": 1330
    },
    {
      "epoch": 1.01,
      "learning_rate": 4.371690620272315e-05,
      "loss": 0.7952,
      "step": 1331
    },
    {
      "epoch": 1.01,
      "learning_rate": 4.371217851739788e-05,
      "loss": 0.8696,
      "step": 1332
    },
    {
      "epoch": 1.01,
      "learning_rate": 4.370745083207262e-05,
      "loss": 0.9264,
      "step": 1333
    },
    {
      "epoch": 1.01,
      "learning_rate": 4.3702723146747354e-05,
      "loss": 0.8891,
      "step": 1334
    },
    {
      "epoch": 1.01,
      "learning_rate": 4.369799546142209e-05,
      "loss": 0.8797,
      "step": 1335
    },
    {
      "epoch": 1.01,
      "learning_rate": 4.3693267776096825e-05,
      "loss": 0.9582,
      "step": 1336
    },
    {
      "epoch": 1.01,
      "learning_rate": 4.3688540090771564e-05,
      "loss": 0.8601,
      "step": 1337
    },
    {
      "epoch": 1.01,
      "learning_rate": 4.3683812405446296e-05,
      "loss": 0.9197,
      "step": 1338
    },
    {
      "epoch": 1.01,
      "learning_rate": 4.3679084720121035e-05,
      "loss": 0.9299,
      "step": 1339
    },
    {
      "epoch": 1.01,
      "learning_rate": 4.367435703479577e-05,
      "loss": 0.8906,
      "step": 1340
    },
    {
      "epoch": 1.01,
      "learning_rate": 4.3669629349470506e-05,
      "loss": 0.8472,
      "step": 1341
    },
    {
      "epoch": 1.01,
      "learning_rate": 4.366490166414524e-05,
      "loss": 0.8138,
      "step": 1342
    },
    {
      "epoch": 1.02,
      "learning_rate": 4.3660173978819976e-05,
      "loss": 0.932,
      "step": 1343
    },
    {
      "epoch": 1.02,
      "learning_rate": 4.365544629349471e-05,
      "loss": 0.9696,
      "step": 1344
    },
    {
      "epoch": 1.02,
      "learning_rate": 4.365071860816945e-05,
      "loss": 0.8577,
      "step": 1345
    },
    {
      "epoch": 1.02,
      "learning_rate": 4.364599092284418e-05,
      "loss": 0.8216,
      "step": 1346
    },
    {
      "epoch": 1.02,
      "learning_rate": 4.364126323751891e-05,
      "loss": 0.7752,
      "step": 1347
    },
    {
      "epoch": 1.02,
      "learning_rate": 4.363653555219365e-05,
      "loss": 0.8389,
      "step": 1348
    },
    {
      "epoch": 1.02,
      "learning_rate": 4.363180786686838e-05,
      "loss": 0.9212,
      "step": 1349
    },
    {
      "epoch": 1.02,
      "learning_rate": 4.3627080181543114e-05,
      "loss": 0.9975,
      "step": 1350
    },
    {
      "epoch": 1.02,
      "learning_rate": 4.362235249621785e-05,
      "loss": 0.9531,
      "step": 1351
    },
    {
      "epoch": 1.02,
      "learning_rate": 4.3617624810892585e-05,
      "loss": 0.8531,
      "step": 1352
    },
    {
      "epoch": 1.02,
      "learning_rate": 4.3612897125567324e-05,
      "loss": 0.8911,
      "step": 1353
    },
    {
      "epoch": 1.02,
      "learning_rate": 4.3608169440242056e-05,
      "loss": 0.7328,
      "step": 1354
    },
    {
      "epoch": 1.02,
      "learning_rate": 4.3603441754916795e-05,
      "loss": 1.0347,
      "step": 1355
    },
    {
      "epoch": 1.03,
      "learning_rate": 4.359871406959153e-05,
      "loss": 0.8704,
      "step": 1356
    },
    {
      "epoch": 1.03,
      "learning_rate": 4.3593986384266266e-05,
      "loss": 0.8498,
      "step": 1357
    },
    {
      "epoch": 1.03,
      "learning_rate": 4.3589258698941e-05,
      "loss": 0.8885,
      "step": 1358
    },
    {
      "epoch": 1.03,
      "learning_rate": 4.358453101361574e-05,
      "loss": 0.8241,
      "step": 1359
    },
    {
      "epoch": 1.03,
      "learning_rate": 4.357980332829047e-05,
      "loss": 1.0142,
      "step": 1360
    },
    {
      "epoch": 1.03,
      "learning_rate": 4.357507564296521e-05,
      "loss": 0.857,
      "step": 1361
    },
    {
      "epoch": 1.03,
      "learning_rate": 4.357034795763994e-05,
      "loss": 0.8673,
      "step": 1362
    },
    {
      "epoch": 1.03,
      "learning_rate": 4.356562027231468e-05,
      "loss": 0.8446,
      "step": 1363
    },
    {
      "epoch": 1.03,
      "learning_rate": 4.356089258698941e-05,
      "loss": 0.803,
      "step": 1364
    },
    {
      "epoch": 1.03,
      "learning_rate": 4.355616490166415e-05,
      "loss": 0.8994,
      "step": 1365
    },
    {
      "epoch": 1.03,
      "learning_rate": 4.355143721633888e-05,
      "loss": 0.7737,
      "step": 1366
    },
    {
      "epoch": 1.03,
      "learning_rate": 4.354670953101362e-05,
      "loss": 0.8799,
      "step": 1367
    },
    {
      "epoch": 1.03,
      "learning_rate": 4.354198184568835e-05,
      "loss": 0.9156,
      "step": 1368
    },
    {
      "epoch": 1.04,
      "learning_rate": 4.353725416036309e-05,
      "loss": 0.9886,
      "step": 1369
    },
    {
      "epoch": 1.04,
      "learning_rate": 4.353252647503782e-05,
      "loss": 0.8198,
      "step": 1370
    },
    {
      "epoch": 1.04,
      "learning_rate": 4.352779878971256e-05,
      "loss": 0.9066,
      "step": 1371
    },
    {
      "epoch": 1.04,
      "learning_rate": 4.3523071104387294e-05,
      "loss": 0.939,
      "step": 1372
    },
    {
      "epoch": 1.04,
      "learning_rate": 4.351834341906203e-05,
      "loss": 0.7408,
      "step": 1373
    },
    {
      "epoch": 1.04,
      "learning_rate": 4.3513615733736765e-05,
      "loss": 0.7962,
      "step": 1374
    },
    {
      "epoch": 1.04,
      "learning_rate": 4.35088880484115e-05,
      "loss": 0.7533,
      "step": 1375
    },
    {
      "epoch": 1.04,
      "learning_rate": 4.3504160363086236e-05,
      "loss": 0.8055,
      "step": 1376
    },
    {
      "epoch": 1.04,
      "learning_rate": 4.349943267776097e-05,
      "loss": 0.9506,
      "step": 1377
    },
    {
      "epoch": 1.04,
      "learning_rate": 4.34947049924357e-05,
      "loss": 0.8286,
      "step": 1378
    },
    {
      "epoch": 1.04,
      "learning_rate": 4.348997730711044e-05,
      "loss": 0.7459,
      "step": 1379
    },
    {
      "epoch": 1.04,
      "learning_rate": 4.348524962178517e-05,
      "loss": 1.015,
      "step": 1380
    },
    {
      "epoch": 1.04,
      "learning_rate": 4.348052193645991e-05,
      "loss": 0.8253,
      "step": 1381
    },
    {
      "epoch": 1.05,
      "learning_rate": 4.347579425113464e-05,
      "loss": 0.8598,
      "step": 1382
    },
    {
      "epoch": 1.05,
      "learning_rate": 4.347106656580938e-05,
      "loss": 0.8823,
      "step": 1383
    },
    {
      "epoch": 1.05,
      "learning_rate": 4.346633888048411e-05,
      "loss": 0.7623,
      "step": 1384
    },
    {
      "epoch": 1.05,
      "learning_rate": 4.346161119515885e-05,
      "loss": 0.7531,
      "step": 1385
    },
    {
      "epoch": 1.05,
      "learning_rate": 4.345688350983359e-05,
      "loss": 0.8568,
      "step": 1386
    },
    {
      "epoch": 1.05,
      "learning_rate": 4.345215582450832e-05,
      "loss": 0.9092,
      "step": 1387
    },
    {
      "epoch": 1.05,
      "learning_rate": 4.344742813918306e-05,
      "loss": 1.021,
      "step": 1388
    },
    {
      "epoch": 1.05,
      "learning_rate": 4.3442700453857793e-05,
      "loss": 0.7631,
      "step": 1389
    },
    {
      "epoch": 1.05,
      "learning_rate": 4.343797276853253e-05,
      "loss": 0.9103,
      "step": 1390
    },
    {
      "epoch": 1.05,
      "learning_rate": 4.3433245083207264e-05,
      "loss": 0.8605,
      "step": 1391
    },
    {
      "epoch": 1.05,
      "learning_rate": 4.3428517397882e-05,
      "loss": 0.89,
      "step": 1392
    },
    {
      "epoch": 1.05,
      "learning_rate": 4.3423789712556735e-05,
      "loss": 0.9027,
      "step": 1393
    },
    {
      "epoch": 1.05,
      "learning_rate": 4.3419062027231474e-05,
      "loss": 0.9731,
      "step": 1394
    },
    {
      "epoch": 1.06,
      "learning_rate": 4.3414334341906206e-05,
      "loss": 0.878,
      "step": 1395
    },
    {
      "epoch": 1.06,
      "learning_rate": 4.3409606656580945e-05,
      "loss": 0.9168,
      "step": 1396
    },
    {
      "epoch": 1.06,
      "learning_rate": 4.340487897125568e-05,
      "loss": 0.7587,
      "step": 1397
    },
    {
      "epoch": 1.06,
      "learning_rate": 4.3400151285930416e-05,
      "loss": 0.9048,
      "step": 1398
    },
    {
      "epoch": 1.06,
      "learning_rate": 4.339542360060515e-05,
      "loss": 0.8379,
      "step": 1399
    },
    {
      "epoch": 1.06,
      "learning_rate": 4.339069591527988e-05,
      "loss": 0.7602,
      "step": 1400
    },
    {
      "epoch": 1.06,
      "learning_rate": 4.338596822995462e-05,
      "loss": 0.9473,
      "step": 1401
    },
    {
      "epoch": 1.06,
      "learning_rate": 4.338124054462935e-05,
      "loss": 0.9421,
      "step": 1402
    },
    {
      "epoch": 1.06,
      "learning_rate": 4.337651285930408e-05,
      "loss": 0.9071,
      "step": 1403
    },
    {
      "epoch": 1.06,
      "learning_rate": 4.337178517397882e-05,
      "loss": 0.8867,
      "step": 1404
    },
    {
      "epoch": 1.06,
      "learning_rate": 4.3367057488653554e-05,
      "loss": 0.9294,
      "step": 1405
    },
    {
      "epoch": 1.06,
      "learning_rate": 4.336232980332829e-05,
      "loss": 0.9821,
      "step": 1406
    },
    {
      "epoch": 1.06,
      "learning_rate": 4.3357602118003025e-05,
      "loss": 0.9136,
      "step": 1407
    },
    {
      "epoch": 1.06,
      "learning_rate": 4.3352874432677764e-05,
      "loss": 0.8982,
      "step": 1408
    },
    {
      "epoch": 1.07,
      "learning_rate": 4.3348146747352496e-05,
      "loss": 0.8049,
      "step": 1409
    },
    {
      "epoch": 1.07,
      "learning_rate": 4.3343419062027234e-05,
      "loss": 0.8598,
      "step": 1410
    },
    {
      "epoch": 1.07,
      "learning_rate": 4.3338691376701967e-05,
      "loss": 0.7809,
      "step": 1411
    },
    {
      "epoch": 1.07,
      "learning_rate": 4.3333963691376705e-05,
      "loss": 0.8908,
      "step": 1412
    },
    {
      "epoch": 1.07,
      "learning_rate": 4.332923600605144e-05,
      "loss": 0.8368,
      "step": 1413
    },
    {
      "epoch": 1.07,
      "learning_rate": 4.3324508320726176e-05,
      "loss": 0.8914,
      "step": 1414
    },
    {
      "epoch": 1.07,
      "learning_rate": 4.331978063540091e-05,
      "loss": 0.8165,
      "step": 1415
    },
    {
      "epoch": 1.07,
      "learning_rate": 4.331505295007565e-05,
      "loss": 0.8113,
      "step": 1416
    },
    {
      "epoch": 1.07,
      "learning_rate": 4.331032526475038e-05,
      "loss": 0.7264,
      "step": 1417
    },
    {
      "epoch": 1.07,
      "learning_rate": 4.330559757942512e-05,
      "loss": 0.785,
      "step": 1418
    },
    {
      "epoch": 1.07,
      "learning_rate": 4.330086989409985e-05,
      "loss": 0.9649,
      "step": 1419
    },
    {
      "epoch": 1.07,
      "learning_rate": 4.329614220877459e-05,
      "loss": 0.7819,
      "step": 1420
    },
    {
      "epoch": 1.07,
      "learning_rate": 4.329141452344932e-05,
      "loss": 0.7547,
      "step": 1421
    },
    {
      "epoch": 1.08,
      "learning_rate": 4.328668683812406e-05,
      "loss": 1.0119,
      "step": 1422
    },
    {
      "epoch": 1.08,
      "learning_rate": 4.328195915279879e-05,
      "loss": 0.806,
      "step": 1423
    },
    {
      "epoch": 1.08,
      "learning_rate": 4.327723146747353e-05,
      "loss": 0.901,
      "step": 1424
    },
    {
      "epoch": 1.08,
      "learning_rate": 4.327250378214826e-05,
      "loss": 0.9185,
      "step": 1425
    },
    {
      "epoch": 1.08,
      "learning_rate": 4.3267776096823e-05,
      "loss": 0.7922,
      "step": 1426
    },
    {
      "epoch": 1.08,
      "learning_rate": 4.3263048411497734e-05,
      "loss": 0.8117,
      "step": 1427
    },
    {
      "epoch": 1.08,
      "learning_rate": 4.3258320726172466e-05,
      "loss": 0.7479,
      "step": 1428
    },
    {
      "epoch": 1.08,
      "learning_rate": 4.3253593040847205e-05,
      "loss": 0.8465,
      "step": 1429
    },
    {
      "epoch": 1.08,
      "learning_rate": 4.324886535552194e-05,
      "loss": 0.7933,
      "step": 1430
    },
    {
      "epoch": 1.08,
      "learning_rate": 4.324413767019667e-05,
      "loss": 0.6972,
      "step": 1431
    },
    {
      "epoch": 1.08,
      "learning_rate": 4.323940998487141e-05,
      "loss": 0.9483,
      "step": 1432
    },
    {
      "epoch": 1.08,
      "learning_rate": 4.323468229954614e-05,
      "loss": 0.7825,
      "step": 1433
    },
    {
      "epoch": 1.08,
      "learning_rate": 4.322995461422088e-05,
      "loss": 0.8995,
      "step": 1434
    },
    {
      "epoch": 1.09,
      "learning_rate": 4.322522692889561e-05,
      "loss": 0.8487,
      "step": 1435
    },
    {
      "epoch": 1.09,
      "learning_rate": 4.322049924357035e-05,
      "loss": 0.8471,
      "step": 1436
    },
    {
      "epoch": 1.09,
      "learning_rate": 4.321577155824508e-05,
      "loss": 0.845,
      "step": 1437
    },
    {
      "epoch": 1.09,
      "learning_rate": 4.321104387291982e-05,
      "loss": 0.8214,
      "step": 1438
    },
    {
      "epoch": 1.09,
      "learning_rate": 4.320631618759455e-05,
      "loss": 0.8422,
      "step": 1439
    },
    {
      "epoch": 1.09,
      "learning_rate": 4.320158850226929e-05,
      "loss": 0.8415,
      "step": 1440
    },
    {
      "epoch": 1.09,
      "learning_rate": 4.319686081694402e-05,
      "loss": 0.9191,
      "step": 1441
    },
    {
      "epoch": 1.09,
      "learning_rate": 4.319213313161876e-05,
      "loss": 0.8765,
      "step": 1442
    },
    {
      "epoch": 1.09,
      "learning_rate": 4.3187405446293494e-05,
      "loss": 0.8604,
      "step": 1443
    },
    {
      "epoch": 1.09,
      "learning_rate": 4.318267776096823e-05,
      "loss": 0.7824,
      "step": 1444
    },
    {
      "epoch": 1.09,
      "learning_rate": 4.317795007564297e-05,
      "loss": 0.8785,
      "step": 1445
    },
    {
      "epoch": 1.09,
      "learning_rate": 4.3173222390317704e-05,
      "loss": 0.917,
      "step": 1446
    },
    {
      "epoch": 1.09,
      "learning_rate": 4.316849470499244e-05,
      "loss": 0.8887,
      "step": 1447
    },
    {
      "epoch": 1.1,
      "learning_rate": 4.3163767019667175e-05,
      "loss": 0.8847,
      "step": 1448
    },
    {
      "epoch": 1.1,
      "learning_rate": 4.3159039334341914e-05,
      "loss": 0.9176,
      "step": 1449
    },
    {
      "epoch": 1.1,
      "learning_rate": 4.3154311649016646e-05,
      "loss": 0.766,
      "step": 1450
    },
    {
      "epoch": 1.1,
      "learning_rate": 4.3149583963691384e-05,
      "loss": 0.8859,
      "step": 1451
    },
    {
      "epoch": 1.1,
      "learning_rate": 4.3144856278366116e-05,
      "loss": 0.9017,
      "step": 1452
    },
    {
      "epoch": 1.1,
      "learning_rate": 4.314012859304085e-05,
      "loss": 0.8623,
      "step": 1453
    },
    {
      "epoch": 1.1,
      "learning_rate": 4.313540090771559e-05,
      "loss": 0.9411,
      "step": 1454
    },
    {
      "epoch": 1.1,
      "learning_rate": 4.313067322239032e-05,
      "loss": 0.8118,
      "step": 1455
    },
    {
      "epoch": 1.1,
      "learning_rate": 4.312594553706505e-05,
      "loss": 0.8643,
      "step": 1456
    },
    {
      "epoch": 1.1,
      "learning_rate": 4.312121785173979e-05,
      "loss": 0.8134,
      "step": 1457
    },
    {
      "epoch": 1.1,
      "learning_rate": 4.311649016641452e-05,
      "loss": 0.8205,
      "step": 1458
    },
    {
      "epoch": 1.1,
      "learning_rate": 4.311176248108926e-05,
      "loss": 0.9299,
      "step": 1459
    },
    {
      "epoch": 1.1,
      "learning_rate": 4.310703479576399e-05,
      "loss": 0.8439,
      "step": 1460
    },
    {
      "epoch": 1.1,
      "learning_rate": 4.310230711043873e-05,
      "loss": 0.8312,
      "step": 1461
    },
    {
      "epoch": 1.11,
      "learning_rate": 4.3097579425113464e-05,
      "loss": 0.7662,
      "step": 1462
    },
    {
      "epoch": 1.11,
      "learning_rate": 4.30928517397882e-05,
      "loss": 0.844,
      "step": 1463
    },
    {
      "epoch": 1.11,
      "learning_rate": 4.3088124054462935e-05,
      "loss": 0.7337,
      "step": 1464
    },
    {
      "epoch": 1.11,
      "learning_rate": 4.3083396369137674e-05,
      "loss": 0.8433,
      "step": 1465
    },
    {
      "epoch": 1.11,
      "learning_rate": 4.3078668683812406e-05,
      "loss": 0.7738,
      "step": 1466
    },
    {
      "epoch": 1.11,
      "learning_rate": 4.3073940998487145e-05,
      "loss": 0.7428,
      "step": 1467
    },
    {
      "epoch": 1.11,
      "learning_rate": 4.306921331316188e-05,
      "loss": 0.8651,
      "step": 1468
    },
    {
      "epoch": 1.11,
      "learning_rate": 4.3064485627836616e-05,
      "loss": 0.8101,
      "step": 1469
    },
    {
      "epoch": 1.11,
      "learning_rate": 4.305975794251135e-05,
      "loss": 0.7759,
      "step": 1470
    },
    {
      "epoch": 1.11,
      "learning_rate": 4.3055030257186087e-05,
      "loss": 0.8624,
      "step": 1471
    },
    {
      "epoch": 1.11,
      "learning_rate": 4.305030257186082e-05,
      "loss": 0.8136,
      "step": 1472
    },
    {
      "epoch": 1.11,
      "learning_rate": 4.304557488653556e-05,
      "loss": 0.6794,
      "step": 1473
    },
    {
      "epoch": 1.11,
      "learning_rate": 4.304084720121029e-05,
      "loss": 0.8695,
      "step": 1474
    },
    {
      "epoch": 1.12,
      "learning_rate": 4.303611951588503e-05,
      "loss": 0.9442,
      "step": 1475
    },
    {
      "epoch": 1.12,
      "learning_rate": 4.303139183055976e-05,
      "loss": 0.8728,
      "step": 1476
    },
    {
      "epoch": 1.12,
      "learning_rate": 4.30266641452345e-05,
      "loss": 0.8692,
      "step": 1477
    },
    {
      "epoch": 1.12,
      "learning_rate": 4.302193645990923e-05,
      "loss": 0.7741,
      "step": 1478
    },
    {
      "epoch": 1.12,
      "learning_rate": 4.301720877458397e-05,
      "loss": 0.855,
      "step": 1479
    },
    {
      "epoch": 1.12,
      "learning_rate": 4.30124810892587e-05,
      "loss": 0.8181,
      "step": 1480
    },
    {
      "epoch": 1.12,
      "learning_rate": 4.3007753403933434e-05,
      "loss": 0.7867,
      "step": 1481
    },
    {
      "epoch": 1.12,
      "learning_rate": 4.300302571860817e-05,
      "loss": 0.8124,
      "step": 1482
    },
    {
      "epoch": 1.12,
      "learning_rate": 4.2998298033282905e-05,
      "loss": 0.8579,
      "step": 1483
    },
    {
      "epoch": 1.12,
      "learning_rate": 4.299357034795764e-05,
      "loss": 0.8431,
      "step": 1484
    },
    {
      "epoch": 1.12,
      "learning_rate": 4.2988842662632376e-05,
      "loss": 0.8276,
      "step": 1485
    },
    {
      "epoch": 1.12,
      "learning_rate": 4.298411497730711e-05,
      "loss": 0.8774,
      "step": 1486
    },
    {
      "epoch": 1.12,
      "learning_rate": 4.297938729198185e-05,
      "loss": 0.8593,
      "step": 1487
    },
    {
      "epoch": 1.13,
      "learning_rate": 4.297465960665658e-05,
      "loss": 0.8257,
      "step": 1488
    },
    {
      "epoch": 1.13,
      "learning_rate": 4.296993192133132e-05,
      "loss": 1.0633,
      "step": 1489
    },
    {
      "epoch": 1.13,
      "learning_rate": 4.296520423600605e-05,
      "loss": 0.8891,
      "step": 1490
    },
    {
      "epoch": 1.13,
      "learning_rate": 4.296047655068079e-05,
      "loss": 0.8403,
      "step": 1491
    },
    {
      "epoch": 1.13,
      "learning_rate": 4.295574886535552e-05,
      "loss": 0.8653,
      "step": 1492
    },
    {
      "epoch": 1.13,
      "learning_rate": 4.295102118003026e-05,
      "loss": 0.8859,
      "step": 1493
    },
    {
      "epoch": 1.13,
      "learning_rate": 4.294629349470499e-05,
      "loss": 0.9025,
      "step": 1494
    },
    {
      "epoch": 1.13,
      "learning_rate": 4.294156580937973e-05,
      "loss": 0.8531,
      "step": 1495
    },
    {
      "epoch": 1.13,
      "learning_rate": 4.293683812405446e-05,
      "loss": 0.9311,
      "step": 1496
    },
    {
      "epoch": 1.13,
      "learning_rate": 4.29321104387292e-05,
      "loss": 0.8676,
      "step": 1497
    },
    {
      "epoch": 1.13,
      "learning_rate": 4.2927382753403934e-05,
      "loss": 0.7758,
      "step": 1498
    },
    {
      "epoch": 1.13,
      "learning_rate": 4.292265506807867e-05,
      "loss": 0.9471,
      "step": 1499
    },
    {
      "epoch": 1.13,
      "learning_rate": 4.2917927382753404e-05,
      "loss": 0.9487,
      "step": 1500
    },
    {
      "epoch": 1.14,
      "learning_rate": 4.291319969742814e-05,
      "loss": 0.7755,
      "step": 1501
    },
    {
      "epoch": 1.14,
      "learning_rate": 4.2908472012102875e-05,
      "loss": 0.9102,
      "step": 1502
    },
    {
      "epoch": 1.14,
      "learning_rate": 4.2903744326777614e-05,
      "loss": 0.9549,
      "step": 1503
    },
    {
      "epoch": 1.14,
      "learning_rate": 4.2899016641452346e-05,
      "loss": 0.8819,
      "step": 1504
    },
    {
      "epoch": 1.14,
      "learning_rate": 4.2894288956127085e-05,
      "loss": 0.8046,
      "step": 1505
    },
    {
      "epoch": 1.14,
      "learning_rate": 4.288956127080182e-05,
      "loss": 0.9226,
      "step": 1506
    },
    {
      "epoch": 1.14,
      "learning_rate": 4.2884833585476556e-05,
      "loss": 0.9341,
      "step": 1507
    },
    {
      "epoch": 1.14,
      "learning_rate": 4.288010590015129e-05,
      "loss": 0.8795,
      "step": 1508
    },
    {
      "epoch": 1.14,
      "learning_rate": 4.287537821482602e-05,
      "loss": 0.8067,
      "step": 1509
    },
    {
      "epoch": 1.14,
      "learning_rate": 4.287065052950076e-05,
      "loss": 0.813,
      "step": 1510
    },
    {
      "epoch": 1.14,
      "learning_rate": 4.286592284417549e-05,
      "loss": 0.788,
      "step": 1511
    },
    {
      "epoch": 1.14,
      "learning_rate": 4.286119515885023e-05,
      "loss": 0.7952,
      "step": 1512
    },
    {
      "epoch": 1.14,
      "learning_rate": 4.285646747352496e-05,
      "loss": 0.8335,
      "step": 1513
    },
    {
      "epoch": 1.15,
      "learning_rate": 4.28517397881997e-05,
      "loss": 1.1374,
      "step": 1514
    },
    {
      "epoch": 1.15,
      "learning_rate": 4.284701210287443e-05,
      "loss": 0.9379,
      "step": 1515
    },
    {
      "epoch": 1.15,
      "learning_rate": 4.284228441754917e-05,
      "loss": 0.8506,
      "step": 1516
    },
    {
      "epoch": 1.15,
      "learning_rate": 4.2837556732223904e-05,
      "loss": 0.8109,
      "step": 1517
    },
    {
      "epoch": 1.15,
      "learning_rate": 4.283282904689864e-05,
      "loss": 0.8367,
      "step": 1518
    },
    {
      "epoch": 1.15,
      "learning_rate": 4.2828101361573375e-05,
      "loss": 0.8209,
      "step": 1519
    },
    {
      "epoch": 1.15,
      "learning_rate": 4.282337367624811e-05,
      "loss": 0.9535,
      "step": 1520
    },
    {
      "epoch": 1.15,
      "learning_rate": 4.2818645990922845e-05,
      "loss": 0.8507,
      "step": 1521
    },
    {
      "epoch": 1.15,
      "learning_rate": 4.2813918305597584e-05,
      "loss": 0.803,
      "step": 1522
    },
    {
      "epoch": 1.15,
      "learning_rate": 4.2809190620272316e-05,
      "loss": 0.8218,
      "step": 1523
    },
    {
      "epoch": 1.15,
      "learning_rate": 4.2804462934947055e-05,
      "loss": 0.9586,
      "step": 1524
    },
    {
      "epoch": 1.15,
      "learning_rate": 4.279973524962179e-05,
      "loss": 0.7745,
      "step": 1525
    },
    {
      "epoch": 1.15,
      "learning_rate": 4.2795007564296526e-05,
      "loss": 0.8738,
      "step": 1526
    },
    {
      "epoch": 1.15,
      "learning_rate": 4.279027987897126e-05,
      "loss": 0.8883,
      "step": 1527
    },
    {
      "epoch": 1.16,
      "learning_rate": 4.2785552193646e-05,
      "loss": 0.8628,
      "step": 1528
    },
    {
      "epoch": 1.16,
      "learning_rate": 4.278082450832073e-05,
      "loss": 0.8503,
      "step": 1529
    },
    {
      "epoch": 1.16,
      "learning_rate": 4.277609682299547e-05,
      "loss": 0.8357,
      "step": 1530
    },
    {
      "epoch": 1.16,
      "learning_rate": 4.27713691376702e-05,
      "loss": 0.9079,
      "step": 1531
    },
    {
      "epoch": 1.16,
      "learning_rate": 4.276664145234494e-05,
      "loss": 0.887,
      "step": 1532
    },
    {
      "epoch": 1.16,
      "learning_rate": 4.276191376701967e-05,
      "loss": 0.7841,
      "step": 1533
    },
    {
      "epoch": 1.16,
      "learning_rate": 4.27571860816944e-05,
      "loss": 0.8119,
      "step": 1534
    },
    {
      "epoch": 1.16,
      "learning_rate": 4.275245839636914e-05,
      "loss": 0.8907,
      "step": 1535
    },
    {
      "epoch": 1.16,
      "learning_rate": 4.2747730711043874e-05,
      "loss": 0.9019,
      "step": 1536
    },
    {
      "epoch": 1.16,
      "learning_rate": 4.2743003025718606e-05,
      "loss": 0.8948,
      "step": 1537
    },
    {
      "epoch": 1.16,
      "learning_rate": 4.2738275340393345e-05,
      "loss": 0.9036,
      "step": 1538
    },
    {
      "epoch": 1.16,
      "learning_rate": 4.273354765506808e-05,
      "loss": 0.8619,
      "step": 1539
    },
    {
      "epoch": 1.16,
      "learning_rate": 4.2728819969742816e-05,
      "loss": 0.9542,
      "step": 1540
    },
    {
      "epoch": 1.17,
      "learning_rate": 4.272409228441755e-05,
      "loss": 0.8055,
      "step": 1541
    },
    {
      "epoch": 1.17,
      "learning_rate": 4.2719364599092286e-05,
      "loss": 0.922,
      "step": 1542
    },
    {
      "epoch": 1.17,
      "learning_rate": 4.271463691376702e-05,
      "loss": 0.8581,
      "step": 1543
    },
    {
      "epoch": 1.17,
      "learning_rate": 4.270990922844176e-05,
      "loss": 0.8089,
      "step": 1544
    },
    {
      "epoch": 1.17,
      "learning_rate": 4.270518154311649e-05,
      "loss": 0.7825,
      "step": 1545
    },
    {
      "epoch": 1.17,
      "learning_rate": 4.270045385779123e-05,
      "loss": 0.815,
      "step": 1546
    },
    {
      "epoch": 1.17,
      "learning_rate": 4.269572617246596e-05,
      "loss": 0.8674,
      "step": 1547
    },
    {
      "epoch": 1.17,
      "learning_rate": 4.26909984871407e-05,
      "loss": 0.8736,
      "step": 1548
    },
    {
      "epoch": 1.17,
      "learning_rate": 4.268627080181543e-05,
      "loss": 0.8222,
      "step": 1549
    },
    {
      "epoch": 1.17,
      "learning_rate": 4.268154311649017e-05,
      "loss": 0.9471,
      "step": 1550
    },
    {
      "epoch": 1.17,
      "learning_rate": 4.26768154311649e-05,
      "loss": 0.8222,
      "step": 1551
    },
    {
      "epoch": 1.17,
      "learning_rate": 4.267208774583964e-05,
      "loss": 0.842,
      "step": 1552
    },
    {
      "epoch": 1.17,
      "learning_rate": 4.266736006051437e-05,
      "loss": 0.8088,
      "step": 1553
    },
    {
      "epoch": 1.18,
      "learning_rate": 4.266263237518911e-05,
      "loss": 0.8122,
      "step": 1554
    },
    {
      "epoch": 1.18,
      "learning_rate": 4.2657904689863844e-05,
      "loss": 0.9702,
      "step": 1555
    },
    {
      "epoch": 1.18,
      "learning_rate": 4.265317700453858e-05,
      "loss": 0.8656,
      "step": 1556
    },
    {
      "epoch": 1.18,
      "learning_rate": 4.2648449319213315e-05,
      "loss": 0.8777,
      "step": 1557
    },
    {
      "epoch": 1.18,
      "learning_rate": 4.2643721633888054e-05,
      "loss": 0.95,
      "step": 1558
    },
    {
      "epoch": 1.18,
      "learning_rate": 4.2638993948562786e-05,
      "loss": 0.7306,
      "step": 1559
    },
    {
      "epoch": 1.18,
      "learning_rate": 4.2634266263237524e-05,
      "loss": 0.7995,
      "step": 1560
    },
    {
      "epoch": 1.18,
      "learning_rate": 4.2629538577912257e-05,
      "loss": 0.7926,
      "step": 1561
    },
    {
      "epoch": 1.18,
      "learning_rate": 4.262481089258699e-05,
      "loss": 0.8442,
      "step": 1562
    },
    {
      "epoch": 1.18,
      "learning_rate": 4.262008320726172e-05,
      "loss": 0.9257,
      "step": 1563
    },
    {
      "epoch": 1.18,
      "learning_rate": 4.261535552193646e-05,
      "loss": 0.8428,
      "step": 1564
    },
    {
      "epoch": 1.18,
      "learning_rate": 4.26106278366112e-05,
      "loss": 0.7743,
      "step": 1565
    },
    {
      "epoch": 1.18,
      "learning_rate": 4.260590015128593e-05,
      "loss": 0.7869,
      "step": 1566
    },
    {
      "epoch": 1.19,
      "learning_rate": 4.260117246596067e-05,
      "loss": 0.8993,
      "step": 1567
    },
    {
      "epoch": 1.19,
      "learning_rate": 4.25964447806354e-05,
      "loss": 0.995,
      "step": 1568
    },
    {
      "epoch": 1.19,
      "learning_rate": 4.259171709531014e-05,
      "loss": 0.8416,
      "step": 1569
    },
    {
      "epoch": 1.19,
      "learning_rate": 4.258698940998487e-05,
      "loss": 0.8034,
      "step": 1570
    },
    {
      "epoch": 1.19,
      "learning_rate": 4.258226172465961e-05,
      "loss": 0.9244,
      "step": 1571
    },
    {
      "epoch": 1.19,
      "learning_rate": 4.257753403933434e-05,
      "loss": 0.8548,
      "step": 1572
    },
    {
      "epoch": 1.19,
      "learning_rate": 4.257280635400908e-05,
      "loss": 0.8532,
      "step": 1573
    },
    {
      "epoch": 1.19,
      "learning_rate": 4.2568078668683814e-05,
      "loss": 0.8529,
      "step": 1574
    },
    {
      "epoch": 1.19,
      "learning_rate": 4.256335098335855e-05,
      "loss": 0.7952,
      "step": 1575
    },
    {
      "epoch": 1.19,
      "learning_rate": 4.2558623298033285e-05,
      "loss": 0.8328,
      "step": 1576
    },
    {
      "epoch": 1.19,
      "learning_rate": 4.2553895612708024e-05,
      "loss": 0.8527,
      "step": 1577
    },
    {
      "epoch": 1.19,
      "learning_rate": 4.2549167927382756e-05,
      "loss": 0.8824,
      "step": 1578
    },
    {
      "epoch": 1.19,
      "learning_rate": 4.2544440242057495e-05,
      "loss": 0.7864,
      "step": 1579
    },
    {
      "epoch": 1.19,
      "learning_rate": 4.253971255673223e-05,
      "loss": 0.7832,
      "step": 1580
    },
    {
      "epoch": 1.2,
      "learning_rate": 4.2534984871406965e-05,
      "loss": 0.8859,
      "step": 1581
    },
    {
      "epoch": 1.2,
      "learning_rate": 4.25302571860817e-05,
      "loss": 0.8357,
      "step": 1582
    },
    {
      "epoch": 1.2,
      "learning_rate": 4.2525529500756436e-05,
      "loss": 0.8737,
      "step": 1583
    },
    {
      "epoch": 1.2,
      "learning_rate": 4.252080181543117e-05,
      "loss": 0.755,
      "step": 1584
    },
    {
      "epoch": 1.2,
      "learning_rate": 4.251607413010591e-05,
      "loss": 0.8423,
      "step": 1585
    },
    {
      "epoch": 1.2,
      "learning_rate": 4.251134644478064e-05,
      "loss": 0.735,
      "step": 1586
    },
    {
      "epoch": 1.2,
      "learning_rate": 4.250661875945537e-05,
      "loss": 0.8226,
      "step": 1587
    },
    {
      "epoch": 1.2,
      "learning_rate": 4.250189107413011e-05,
      "loss": 0.998,
      "step": 1588
    },
    {
      "epoch": 1.2,
      "learning_rate": 4.249716338880484e-05,
      "loss": 0.9116,
      "step": 1589
    },
    {
      "epoch": 1.2,
      "learning_rate": 4.2492435703479574e-05,
      "loss": 0.9074,
      "step": 1590
    },
    {
      "epoch": 1.2,
      "learning_rate": 4.248770801815431e-05,
      "loss": 0.7349,
      "step": 1591
    },
    {
      "epoch": 1.2,
      "learning_rate": 4.2482980332829045e-05,
      "loss": 0.9172,
      "step": 1592
    },
    {
      "epoch": 1.2,
      "learning_rate": 4.2478252647503784e-05,
      "loss": 0.9511,
      "step": 1593
    },
    {
      "epoch": 1.21,
      "learning_rate": 4.2473524962178516e-05,
      "loss": 0.8758,
      "step": 1594
    },
    {
      "epoch": 1.21,
      "learning_rate": 4.2468797276853255e-05,
      "loss": 0.8271,
      "step": 1595
    },
    {
      "epoch": 1.21,
      "learning_rate": 4.246406959152799e-05,
      "loss": 0.7903,
      "step": 1596
    },
    {
      "epoch": 1.21,
      "learning_rate": 4.2459341906202726e-05,
      "loss": 0.9691,
      "step": 1597
    },
    {
      "epoch": 1.21,
      "learning_rate": 4.245461422087746e-05,
      "loss": 0.9107,
      "step": 1598
    },
    {
      "epoch": 1.21,
      "learning_rate": 4.24498865355522e-05,
      "loss": 0.9051,
      "step": 1599
    },
    {
      "epoch": 1.21,
      "learning_rate": 4.244515885022693e-05,
      "loss": 0.8284,
      "step": 1600
    },
    {
      "epoch": 1.21,
      "learning_rate": 4.244043116490167e-05,
      "loss": 0.9758,
      "step": 1601
    },
    {
      "epoch": 1.21,
      "learning_rate": 4.24357034795764e-05,
      "loss": 0.7196,
      "step": 1602
    },
    {
      "epoch": 1.21,
      "learning_rate": 4.243097579425114e-05,
      "loss": 0.9545,
      "step": 1603
    },
    {
      "epoch": 1.21,
      "learning_rate": 4.242624810892587e-05,
      "loss": 0.9131,
      "step": 1604
    },
    {
      "epoch": 1.21,
      "learning_rate": 4.242152042360061e-05,
      "loss": 0.7639,
      "step": 1605
    },
    {
      "epoch": 1.21,
      "learning_rate": 4.241679273827534e-05,
      "loss": 0.7627,
      "step": 1606
    },
    {
      "epoch": 1.22,
      "learning_rate": 4.241206505295008e-05,
      "loss": 0.8159,
      "step": 1607
    },
    {
      "epoch": 1.22,
      "learning_rate": 4.240733736762481e-05,
      "loss": 0.9316,
      "step": 1608
    },
    {
      "epoch": 1.22,
      "learning_rate": 4.240260968229955e-05,
      "loss": 0.848,
      "step": 1609
    },
    {
      "epoch": 1.22,
      "learning_rate": 4.239788199697428e-05,
      "loss": 0.8091,
      "step": 1610
    },
    {
      "epoch": 1.22,
      "learning_rate": 4.239315431164902e-05,
      "loss": 0.9495,
      "step": 1611
    },
    {
      "epoch": 1.22,
      "learning_rate": 4.2388426626323754e-05,
      "loss": 0.8154,
      "step": 1612
    },
    {
      "epoch": 1.22,
      "learning_rate": 4.238369894099849e-05,
      "loss": 0.9062,
      "step": 1613
    },
    {
      "epoch": 1.22,
      "learning_rate": 4.2378971255673225e-05,
      "loss": 0.9265,
      "step": 1614
    },
    {
      "epoch": 1.22,
      "learning_rate": 4.237424357034796e-05,
      "loss": 0.954,
      "step": 1615
    },
    {
      "epoch": 1.22,
      "learning_rate": 4.236951588502269e-05,
      "loss": 0.9568,
      "step": 1616
    },
    {
      "epoch": 1.22,
      "learning_rate": 4.236478819969743e-05,
      "loss": 0.9024,
      "step": 1617
    },
    {
      "epoch": 1.22,
      "learning_rate": 4.236006051437216e-05,
      "loss": 0.8591,
      "step": 1618
    },
    {
      "epoch": 1.22,
      "learning_rate": 4.23553328290469e-05,
      "loss": 0.9678,
      "step": 1619
    },
    {
      "epoch": 1.23,
      "learning_rate": 4.235060514372163e-05,
      "loss": 0.8106,
      "step": 1620
    },
    {
      "epoch": 1.23,
      "learning_rate": 4.234587745839637e-05,
      "loss": 0.8483,
      "step": 1621
    },
    {
      "epoch": 1.23,
      "learning_rate": 4.23411497730711e-05,
      "loss": 0.8708,
      "step": 1622
    },
    {
      "epoch": 1.23,
      "learning_rate": 4.233642208774584e-05,
      "loss": 0.7964,
      "step": 1623
    },
    {
      "epoch": 1.23,
      "learning_rate": 4.233169440242057e-05,
      "loss": 0.8709,
      "step": 1624
    },
    {
      "epoch": 1.23,
      "learning_rate": 4.232696671709531e-05,
      "loss": 0.8516,
      "step": 1625
    },
    {
      "epoch": 1.23,
      "learning_rate": 4.232223903177005e-05,
      "loss": 0.9645,
      "step": 1626
    },
    {
      "epoch": 1.23,
      "learning_rate": 4.231751134644478e-05,
      "loss": 0.815,
      "step": 1627
    },
    {
      "epoch": 1.23,
      "learning_rate": 4.231278366111952e-05,
      "loss": 0.8976,
      "step": 1628
    },
    {
      "epoch": 1.23,
      "learning_rate": 4.2308055975794253e-05,
      "loss": 0.9141,
      "step": 1629
    },
    {
      "epoch": 1.23,
      "learning_rate": 4.230332829046899e-05,
      "loss": 0.7924,
      "step": 1630
    },
    {
      "epoch": 1.23,
      "learning_rate": 4.2298600605143724e-05,
      "loss": 0.8386,
      "step": 1631
    },
    {
      "epoch": 1.23,
      "learning_rate": 4.229387291981846e-05,
      "loss": 1.0109,
      "step": 1632
    },
    {
      "epoch": 1.24,
      "learning_rate": 4.2289145234493195e-05,
      "loss": 0.8562,
      "step": 1633
    },
    {
      "epoch": 1.24,
      "learning_rate": 4.2284417549167934e-05,
      "loss": 0.729,
      "step": 1634
    },
    {
      "epoch": 1.24,
      "learning_rate": 4.2279689863842666e-05,
      "loss": 0.8945,
      "step": 1635
    },
    {
      "epoch": 1.24,
      "learning_rate": 4.2274962178517405e-05,
      "loss": 0.8084,
      "step": 1636
    },
    {
      "epoch": 1.24,
      "learning_rate": 4.227023449319214e-05,
      "loss": 0.826,
      "step": 1637
    },
    {
      "epoch": 1.24,
      "learning_rate": 4.2265506807866876e-05,
      "loss": 0.9356,
      "step": 1638
    },
    {
      "epoch": 1.24,
      "learning_rate": 4.226077912254161e-05,
      "loss": 0.8225,
      "step": 1639
    },
    {
      "epoch": 1.24,
      "learning_rate": 4.225605143721634e-05,
      "loss": 0.7311,
      "step": 1640
    },
    {
      "epoch": 1.24,
      "learning_rate": 4.225132375189108e-05,
      "loss": 0.9405,
      "step": 1641
    },
    {
      "epoch": 1.24,
      "learning_rate": 4.224659606656581e-05,
      "loss": 0.8766,
      "step": 1642
    },
    {
      "epoch": 1.24,
      "learning_rate": 4.224186838124054e-05,
      "loss": 0.788,
      "step": 1643
    },
    {
      "epoch": 1.24,
      "learning_rate": 4.223714069591528e-05,
      "loss": 0.9193,
      "step": 1644
    },
    {
      "epoch": 1.24,
      "learning_rate": 4.2232413010590014e-05,
      "loss": 0.9221,
      "step": 1645
    },
    {
      "epoch": 1.24,
      "learning_rate": 4.222768532526475e-05,
      "loss": 0.7396,
      "step": 1646
    },
    {
      "epoch": 1.25,
      "learning_rate": 4.2222957639939485e-05,
      "loss": 0.9786,
      "step": 1647
    },
    {
      "epoch": 1.25,
      "learning_rate": 4.2218229954614224e-05,
      "loss": 0.8411,
      "step": 1648
    },
    {
      "epoch": 1.25,
      "learning_rate": 4.2213502269288956e-05,
      "loss": 0.8706,
      "step": 1649
    },
    {
      "epoch": 1.25,
      "learning_rate": 4.2208774583963694e-05,
      "loss": 0.9231,
      "step": 1650
    },
    {
      "epoch": 1.25,
      "learning_rate": 4.2204046898638426e-05,
      "loss": 1.0235,
      "step": 1651
    },
    {
      "epoch": 1.25,
      "learning_rate": 4.2199319213313165e-05,
      "loss": 0.9003,
      "step": 1652
    },
    {
      "epoch": 1.25,
      "learning_rate": 4.21945915279879e-05,
      "loss": 0.8166,
      "step": 1653
    },
    {
      "epoch": 1.25,
      "learning_rate": 4.2189863842662636e-05,
      "loss": 0.9454,
      "step": 1654
    },
    {
      "epoch": 1.25,
      "learning_rate": 4.218513615733737e-05,
      "loss": 0.9311,
      "step": 1655
    },
    {
      "epoch": 1.25,
      "learning_rate": 4.218040847201211e-05,
      "loss": 0.8229,
      "step": 1656
    },
    {
      "epoch": 1.25,
      "learning_rate": 4.217568078668684e-05,
      "loss": 0.8721,
      "step": 1657
    },
    {
      "epoch": 1.25,
      "learning_rate": 4.217095310136158e-05,
      "loss": 0.7638,
      "step": 1658
    },
    {
      "epoch": 1.25,
      "learning_rate": 4.216622541603631e-05,
      "loss": 0.6939,
      "step": 1659
    },
    {
      "epoch": 1.26,
      "learning_rate": 4.216149773071105e-05,
      "loss": 0.777,
      "step": 1660
    },
    {
      "epoch": 1.26,
      "learning_rate": 4.215677004538578e-05,
      "loss": 0.8443,
      "step": 1661
    },
    {
      "epoch": 1.26,
      "learning_rate": 4.215204236006052e-05,
      "loss": 0.7513,
      "step": 1662
    },
    {
      "epoch": 1.26,
      "learning_rate": 4.214731467473525e-05,
      "loss": 0.8369,
      "step": 1663
    },
    {
      "epoch": 1.26,
      "learning_rate": 4.214258698940999e-05,
      "loss": 0.9308,
      "step": 1664
    },
    {
      "epoch": 1.26,
      "learning_rate": 4.213785930408472e-05,
      "loss": 0.8457,
      "step": 1665
    },
    {
      "epoch": 1.26,
      "learning_rate": 4.213313161875946e-05,
      "loss": 0.8494,
      "step": 1666
    },
    {
      "epoch": 1.26,
      "learning_rate": 4.2128403933434194e-05,
      "loss": 0.9063,
      "step": 1667
    },
    {
      "epoch": 1.26,
      "learning_rate": 4.2123676248108926e-05,
      "loss": 0.8056,
      "step": 1668
    },
    {
      "epoch": 1.26,
      "learning_rate": 4.211894856278366e-05,
      "loss": 0.9441,
      "step": 1669
    },
    {
      "epoch": 1.26,
      "learning_rate": 4.2114220877458397e-05,
      "loss": 0.887,
      "step": 1670
    },
    {
      "epoch": 1.26,
      "learning_rate": 4.210949319213313e-05,
      "loss": 0.9082,
      "step": 1671
    },
    {
      "epoch": 1.26,
      "learning_rate": 4.210476550680787e-05,
      "loss": 0.783,
      "step": 1672
    },
    {
      "epoch": 1.27,
      "learning_rate": 4.21000378214826e-05,
      "loss": 0.7935,
      "step": 1673
    },
    {
      "epoch": 1.27,
      "learning_rate": 4.209531013615734e-05,
      "loss": 0.8344,
      "step": 1674
    },
    {
      "epoch": 1.27,
      "learning_rate": 4.209058245083207e-05,
      "loss": 1.0467,
      "step": 1675
    },
    {
      "epoch": 1.27,
      "learning_rate": 4.208585476550681e-05,
      "loss": 0.9096,
      "step": 1676
    },
    {
      "epoch": 1.27,
      "learning_rate": 4.208112708018154e-05,
      "loss": 0.9086,
      "step": 1677
    },
    {
      "epoch": 1.27,
      "learning_rate": 4.207639939485628e-05,
      "loss": 0.9431,
      "step": 1678
    },
    {
      "epoch": 1.27,
      "learning_rate": 4.207167170953101e-05,
      "loss": 0.9967,
      "step": 1679
    },
    {
      "epoch": 1.27,
      "learning_rate": 4.206694402420575e-05,
      "loss": 1.0193,
      "step": 1680
    },
    {
      "epoch": 1.27,
      "learning_rate": 4.206221633888048e-05,
      "loss": 0.8848,
      "step": 1681
    },
    {
      "epoch": 1.27,
      "learning_rate": 4.205748865355522e-05,
      "loss": 0.7985,
      "step": 1682
    },
    {
      "epoch": 1.27,
      "learning_rate": 4.2052760968229954e-05,
      "loss": 0.8762,
      "step": 1683
    },
    {
      "epoch": 1.27,
      "learning_rate": 4.204803328290469e-05,
      "loss": 0.9228,
      "step": 1684
    },
    {
      "epoch": 1.27,
      "learning_rate": 4.204330559757943e-05,
      "loss": 0.8014,
      "step": 1685
    },
    {
      "epoch": 1.28,
      "learning_rate": 4.2038577912254164e-05,
      "loss": 0.9721,
      "step": 1686
    },
    {
      "epoch": 1.28,
      "learning_rate": 4.20338502269289e-05,
      "loss": 0.8018,
      "step": 1687
    },
    {
      "epoch": 1.28,
      "learning_rate": 4.2029122541603635e-05,
      "loss": 0.9443,
      "step": 1688
    },
    {
      "epoch": 1.28,
      "learning_rate": 4.2024394856278373e-05,
      "loss": 0.9255,
      "step": 1689
    },
    {
      "epoch": 1.28,
      "learning_rate": 4.2019667170953106e-05,
      "loss": 0.8462,
      "step": 1690
    },
    {
      "epoch": 1.28,
      "learning_rate": 4.2014939485627844e-05,
      "loss": 0.8646,
      "step": 1691
    },
    {
      "epoch": 1.28,
      "learning_rate": 4.2010211800302576e-05,
      "loss": 0.7905,
      "step": 1692
    },
    {
      "epoch": 1.28,
      "learning_rate": 4.200548411497731e-05,
      "loss": 0.7606,
      "step": 1693
    },
    {
      "epoch": 1.28,
      "learning_rate": 4.200075642965205e-05,
      "loss": 0.9109,
      "step": 1694
    },
    {
      "epoch": 1.28,
      "learning_rate": 4.199602874432678e-05,
      "loss": 0.8445,
      "step": 1695
    },
    {
      "epoch": 1.28,
      "learning_rate": 4.199130105900151e-05,
      "loss": 0.8244,
      "step": 1696
    },
    {
      "epoch": 1.28,
      "learning_rate": 4.198657337367625e-05,
      "loss": 0.9446,
      "step": 1697
    },
    {
      "epoch": 1.28,
      "learning_rate": 4.198184568835098e-05,
      "loss": 0.9808,
      "step": 1698
    },
    {
      "epoch": 1.28,
      "learning_rate": 4.197711800302572e-05,
      "loss": 0.7856,
      "step": 1699
    },
    {
      "epoch": 1.29,
      "learning_rate": 4.197239031770045e-05,
      "loss": 0.8073,
      "step": 1700
    },
    {
      "epoch": 1.29,
      "learning_rate": 4.196766263237519e-05,
      "loss": 1.0676,
      "step": 1701
    },
    {
      "epoch": 1.29,
      "learning_rate": 4.1962934947049924e-05,
      "loss": 0.9421,
      "step": 1702
    },
    {
      "epoch": 1.29,
      "learning_rate": 4.195820726172466e-05,
      "loss": 0.9531,
      "step": 1703
    },
    {
      "epoch": 1.29,
      "learning_rate": 4.1953479576399395e-05,
      "loss": 0.8569,
      "step": 1704
    },
    {
      "epoch": 1.29,
      "learning_rate": 4.1948751891074134e-05,
      "loss": 0.9385,
      "step": 1705
    },
    {
      "epoch": 1.29,
      "learning_rate": 4.1944024205748866e-05,
      "loss": 0.9515,
      "step": 1706
    },
    {
      "epoch": 1.29,
      "learning_rate": 4.1939296520423605e-05,
      "loss": 0.7696,
      "step": 1707
    },
    {
      "epoch": 1.29,
      "learning_rate": 4.193456883509834e-05,
      "loss": 0.8316,
      "step": 1708
    },
    {
      "epoch": 1.29,
      "learning_rate": 4.1929841149773076e-05,
      "loss": 0.8938,
      "step": 1709
    },
    {
      "epoch": 1.29,
      "learning_rate": 4.192511346444781e-05,
      "loss": 0.8477,
      "step": 1710
    },
    {
      "epoch": 1.29,
      "learning_rate": 4.1920385779122547e-05,
      "loss": 0.8863,
      "step": 1711
    },
    {
      "epoch": 1.29,
      "learning_rate": 4.191565809379728e-05,
      "loss": 0.7614,
      "step": 1712
    },
    {
      "epoch": 1.3,
      "learning_rate": 4.191093040847202e-05,
      "loss": 0.9538,
      "step": 1713
    },
    {
      "epoch": 1.3,
      "learning_rate": 4.190620272314675e-05,
      "loss": 0.8395,
      "step": 1714
    },
    {
      "epoch": 1.3,
      "learning_rate": 4.190147503782149e-05,
      "loss": 0.7395,
      "step": 1715
    },
    {
      "epoch": 1.3,
      "learning_rate": 4.189674735249622e-05,
      "loss": 0.8881,
      "step": 1716
    },
    {
      "epoch": 1.3,
      "learning_rate": 4.189201966717096e-05,
      "loss": 0.8481,
      "step": 1717
    },
    {
      "epoch": 1.3,
      "learning_rate": 4.188729198184569e-05,
      "loss": 0.9458,
      "step": 1718
    },
    {
      "epoch": 1.3,
      "learning_rate": 4.188256429652043e-05,
      "loss": 0.889,
      "step": 1719
    },
    {
      "epoch": 1.3,
      "learning_rate": 4.187783661119516e-05,
      "loss": 0.8094,
      "step": 1720
    },
    {
      "epoch": 1.3,
      "learning_rate": 4.1873108925869894e-05,
      "loss": 0.8775,
      "step": 1721
    },
    {
      "epoch": 1.3,
      "learning_rate": 4.1868381240544626e-05,
      "loss": 0.8589,
      "step": 1722
    },
    {
      "epoch": 1.3,
      "learning_rate": 4.1863653555219365e-05,
      "loss": 0.7877,
      "step": 1723
    },
    {
      "epoch": 1.3,
      "learning_rate": 4.18589258698941e-05,
      "loss": 0.876,
      "step": 1724
    },
    {
      "epoch": 1.3,
      "learning_rate": 4.1854198184568836e-05,
      "loss": 0.7768,
      "step": 1725
    },
    {
      "epoch": 1.31,
      "learning_rate": 4.184947049924357e-05,
      "loss": 0.9494,
      "step": 1726
    },
    {
      "epoch": 1.31,
      "learning_rate": 4.184474281391831e-05,
      "loss": 0.8285,
      "step": 1727
    },
    {
      "epoch": 1.31,
      "learning_rate": 4.184001512859304e-05,
      "loss": 0.9739,
      "step": 1728
    },
    {
      "epoch": 1.31,
      "learning_rate": 4.183528744326778e-05,
      "loss": 0.8771,
      "step": 1729
    },
    {
      "epoch": 1.31,
      "learning_rate": 4.183055975794251e-05,
      "loss": 0.88,
      "step": 1730
    },
    {
      "epoch": 1.31,
      "learning_rate": 4.182583207261725e-05,
      "loss": 0.8839,
      "step": 1731
    },
    {
      "epoch": 1.31,
      "learning_rate": 4.182110438729198e-05,
      "loss": 0.7774,
      "step": 1732
    },
    {
      "epoch": 1.31,
      "learning_rate": 4.181637670196672e-05,
      "loss": 0.8566,
      "step": 1733
    },
    {
      "epoch": 1.31,
      "learning_rate": 4.181164901664145e-05,
      "loss": 0.9113,
      "step": 1734
    },
    {
      "epoch": 1.31,
      "learning_rate": 4.180692133131619e-05,
      "loss": 0.9548,
      "step": 1735
    },
    {
      "epoch": 1.31,
      "learning_rate": 4.180219364599092e-05,
      "loss": 0.8681,
      "step": 1736
    },
    {
      "epoch": 1.31,
      "learning_rate": 4.179746596066566e-05,
      "loss": 0.8832,
      "step": 1737
    },
    {
      "epoch": 1.31,
      "learning_rate": 4.1792738275340393e-05,
      "loss": 0.8257,
      "step": 1738
    },
    {
      "epoch": 1.32,
      "learning_rate": 4.178801059001513e-05,
      "loss": 1.0088,
      "step": 1739
    },
    {
      "epoch": 1.32,
      "learning_rate": 4.1783282904689864e-05,
      "loss": 0.8771,
      "step": 1740
    },
    {
      "epoch": 1.32,
      "learning_rate": 4.17785552193646e-05,
      "loss": 0.8337,
      "step": 1741
    },
    {
      "epoch": 1.32,
      "learning_rate": 4.1773827534039335e-05,
      "loss": 0.8956,
      "step": 1742
    },
    {
      "epoch": 1.32,
      "learning_rate": 4.1769099848714074e-05,
      "loss": 0.7947,
      "step": 1743
    },
    {
      "epoch": 1.32,
      "learning_rate": 4.1764372163388806e-05,
      "loss": 0.8388,
      "step": 1744
    },
    {
      "epoch": 1.32,
      "learning_rate": 4.1759644478063545e-05,
      "loss": 0.8279,
      "step": 1745
    },
    {
      "epoch": 1.32,
      "learning_rate": 4.175491679273828e-05,
      "loss": 0.7525,
      "step": 1746
    },
    {
      "epoch": 1.32,
      "learning_rate": 4.1750189107413016e-05,
      "loss": 0.8997,
      "step": 1747
    },
    {
      "epoch": 1.32,
      "learning_rate": 4.174546142208775e-05,
      "loss": 0.8499,
      "step": 1748
    },
    {
      "epoch": 1.32,
      "learning_rate": 4.174073373676248e-05,
      "loss": 0.9435,
      "step": 1749
    },
    {
      "epoch": 1.32,
      "learning_rate": 4.173600605143722e-05,
      "loss": 0.9114,
      "step": 1750
    },
    {
      "epoch": 1.32,
      "learning_rate": 4.173127836611195e-05,
      "loss": 1.0919,
      "step": 1751
    },
    {
      "epoch": 1.33,
      "learning_rate": 4.172655068078669e-05,
      "loss": 0.7866,
      "step": 1752
    },
    {
      "epoch": 1.33,
      "learning_rate": 4.172182299546142e-05,
      "loss": 0.8788,
      "step": 1753
    },
    {
      "epoch": 1.33,
      "learning_rate": 4.171709531013616e-05,
      "loss": 0.7924,
      "step": 1754
    },
    {
      "epoch": 1.33,
      "learning_rate": 4.171236762481089e-05,
      "loss": 0.8005,
      "step": 1755
    },
    {
      "epoch": 1.33,
      "learning_rate": 4.170763993948563e-05,
      "loss": 0.9331,
      "step": 1756
    },
    {
      "epoch": 1.33,
      "learning_rate": 4.1702912254160364e-05,
      "loss": 0.8057,
      "step": 1757
    },
    {
      "epoch": 1.33,
      "learning_rate": 4.16981845688351e-05,
      "loss": 0.8805,
      "step": 1758
    },
    {
      "epoch": 1.33,
      "learning_rate": 4.1693456883509834e-05,
      "loss": 0.9478,
      "step": 1759
    },
    {
      "epoch": 1.33,
      "learning_rate": 4.168872919818457e-05,
      "loss": 0.9375,
      "step": 1760
    },
    {
      "epoch": 1.33,
      "learning_rate": 4.1684001512859305e-05,
      "loss": 0.9896,
      "step": 1761
    },
    {
      "epoch": 1.33,
      "learning_rate": 4.1679273827534044e-05,
      "loss": 0.8581,
      "step": 1762
    },
    {
      "epoch": 1.33,
      "learning_rate": 4.1674546142208776e-05,
      "loss": 0.7759,
      "step": 1763
    },
    {
      "epoch": 1.33,
      "learning_rate": 4.1669818456883515e-05,
      "loss": 0.9268,
      "step": 1764
    },
    {
      "epoch": 1.33,
      "learning_rate": 4.166509077155825e-05,
      "loss": 0.7976,
      "step": 1765
    },
    {
      "epoch": 1.34,
      "learning_rate": 4.1660363086232986e-05,
      "loss": 0.9293,
      "step": 1766
    },
    {
      "epoch": 1.34,
      "learning_rate": 4.165563540090772e-05,
      "loss": 0.9192,
      "step": 1767
    },
    {
      "epoch": 1.34,
      "learning_rate": 4.165090771558246e-05,
      "loss": 0.671,
      "step": 1768
    },
    {
      "epoch": 1.34,
      "learning_rate": 4.164618003025719e-05,
      "loss": 0.8176,
      "step": 1769
    },
    {
      "epoch": 1.34,
      "learning_rate": 4.164145234493193e-05,
      "loss": 0.9001,
      "step": 1770
    },
    {
      "epoch": 1.34,
      "learning_rate": 4.163672465960666e-05,
      "loss": 0.9391,
      "step": 1771
    },
    {
      "epoch": 1.34,
      "learning_rate": 4.16319969742814e-05,
      "loss": 0.789,
      "step": 1772
    },
    {
      "epoch": 1.34,
      "learning_rate": 4.162726928895613e-05,
      "loss": 0.8756,
      "step": 1773
    },
    {
      "epoch": 1.34,
      "learning_rate": 4.162254160363086e-05,
      "loss": 0.8055,
      "step": 1774
    },
    {
      "epoch": 1.34,
      "learning_rate": 4.1617813918305595e-05,
      "loss": 0.7769,
      "step": 1775
    },
    {
      "epoch": 1.34,
      "learning_rate": 4.1613086232980334e-05,
      "loss": 0.7872,
      "step": 1776
    },
    {
      "epoch": 1.34,
      "learning_rate": 4.1608358547655066e-05,
      "loss": 0.8305,
      "step": 1777
    },
    {
      "epoch": 1.34,
      "learning_rate": 4.1603630862329805e-05,
      "loss": 0.9093,
      "step": 1778
    },
    {
      "epoch": 1.35,
      "learning_rate": 4.159890317700454e-05,
      "loss": 0.9944,
      "step": 1779
    },
    {
      "epoch": 1.35,
      "learning_rate": 4.1594175491679275e-05,
      "loss": 0.8518,
      "step": 1780
    },
    {
      "epoch": 1.35,
      "learning_rate": 4.158944780635401e-05,
      "loss": 0.8338,
      "step": 1781
    },
    {
      "epoch": 1.35,
      "learning_rate": 4.1584720121028746e-05,
      "loss": 0.8605,
      "step": 1782
    },
    {
      "epoch": 1.35,
      "learning_rate": 4.157999243570348e-05,
      "loss": 0.9801,
      "step": 1783
    },
    {
      "epoch": 1.35,
      "learning_rate": 4.157526475037822e-05,
      "loss": 0.7776,
      "step": 1784
    },
    {
      "epoch": 1.35,
      "learning_rate": 4.157053706505295e-05,
      "loss": 0.8093,
      "step": 1785
    },
    {
      "epoch": 1.35,
      "learning_rate": 4.156580937972769e-05,
      "loss": 0.8692,
      "step": 1786
    },
    {
      "epoch": 1.35,
      "learning_rate": 4.156108169440242e-05,
      "loss": 0.9492,
      "step": 1787
    },
    {
      "epoch": 1.35,
      "learning_rate": 4.155635400907716e-05,
      "loss": 0.8049,
      "step": 1788
    },
    {
      "epoch": 1.35,
      "learning_rate": 4.155162632375189e-05,
      "loss": 0.7929,
      "step": 1789
    },
    {
      "epoch": 1.35,
      "learning_rate": 4.154689863842663e-05,
      "loss": 1.0079,
      "step": 1790
    },
    {
      "epoch": 1.35,
      "learning_rate": 4.154217095310136e-05,
      "loss": 0.9516,
      "step": 1791
    },
    {
      "epoch": 1.36,
      "learning_rate": 4.15374432677761e-05,
      "loss": 0.8518,
      "step": 1792
    },
    {
      "epoch": 1.36,
      "learning_rate": 4.153271558245083e-05,
      "loss": 0.8686,
      "step": 1793
    },
    {
      "epoch": 1.36,
      "learning_rate": 4.152798789712557e-05,
      "loss": 0.8162,
      "step": 1794
    },
    {
      "epoch": 1.36,
      "learning_rate": 4.1523260211800304e-05,
      "loss": 0.8069,
      "step": 1795
    },
    {
      "epoch": 1.36,
      "learning_rate": 4.151853252647504e-05,
      "loss": 0.9564,
      "step": 1796
    },
    {
      "epoch": 1.36,
      "learning_rate": 4.1513804841149775e-05,
      "loss": 0.9981,
      "step": 1797
    },
    {
      "epoch": 1.36,
      "learning_rate": 4.1509077155824514e-05,
      "loss": 0.7983,
      "step": 1798
    },
    {
      "epoch": 1.36,
      "learning_rate": 4.1504349470499246e-05,
      "loss": 0.9547,
      "step": 1799
    },
    {
      "epoch": 1.36,
      "learning_rate": 4.149962178517398e-05,
      "loss": 0.915,
      "step": 1800
    },
    {
      "epoch": 1.36,
      "learning_rate": 4.1494894099848716e-05,
      "loss": 0.8171,
      "step": 1801
    },
    {
      "epoch": 1.36,
      "learning_rate": 4.149016641452345e-05,
      "loss": 0.9512,
      "step": 1802
    },
    {
      "epoch": 1.36,
      "learning_rate": 4.148543872919818e-05,
      "loss": 0.8199,
      "step": 1803
    },
    {
      "epoch": 1.36,
      "learning_rate": 4.148071104387292e-05,
      "loss": 0.7785,
      "step": 1804
    },
    {
      "epoch": 1.37,
      "learning_rate": 4.147598335854766e-05,
      "loss": 0.8781,
      "step": 1805
    },
    {
      "epoch": 1.37,
      "learning_rate": 4.147125567322239e-05,
      "loss": 0.8464,
      "step": 1806
    },
    {
      "epoch": 1.37,
      "learning_rate": 4.146652798789713e-05,
      "loss": 0.8284,
      "step": 1807
    },
    {
      "epoch": 1.37,
      "learning_rate": 4.146180030257186e-05,
      "loss": 0.9802,
      "step": 1808
    },
    {
      "epoch": 1.37,
      "learning_rate": 4.14570726172466e-05,
      "loss": 0.8728,
      "step": 1809
    },
    {
      "epoch": 1.37,
      "learning_rate": 4.145234493192133e-05,
      "loss": 0.8611,
      "step": 1810
    },
    {
      "epoch": 1.37,
      "learning_rate": 4.144761724659607e-05,
      "loss": 0.981,
      "step": 1811
    },
    {
      "epoch": 1.37,
      "learning_rate": 4.14428895612708e-05,
      "loss": 0.7918,
      "step": 1812
    },
    {
      "epoch": 1.37,
      "learning_rate": 4.143816187594554e-05,
      "loss": 0.8416,
      "step": 1813
    },
    {
      "epoch": 1.37,
      "learning_rate": 4.1433434190620274e-05,
      "loss": 0.9431,
      "step": 1814
    },
    {
      "epoch": 1.37,
      "learning_rate": 4.142870650529501e-05,
      "loss": 0.8068,
      "step": 1815
    },
    {
      "epoch": 1.37,
      "learning_rate": 4.1423978819969745e-05,
      "loss": 0.8667,
      "step": 1816
    },
    {
      "epoch": 1.37,
      "learning_rate": 4.1419251134644484e-05,
      "loss": 1.0549,
      "step": 1817
    },
    {
      "epoch": 1.37,
      "learning_rate": 4.1414523449319216e-05,
      "loss": 0.8903,
      "step": 1818
    },
    {
      "epoch": 1.38,
      "learning_rate": 4.1409795763993955e-05,
      "loss": 0.8751,
      "step": 1819
    },
    {
      "epoch": 1.38,
      "learning_rate": 4.1405068078668687e-05,
      "loss": 0.8569,
      "step": 1820
    },
    {
      "epoch": 1.38,
      "learning_rate": 4.1400340393343425e-05,
      "loss": 0.832,
      "step": 1821
    },
    {
      "epoch": 1.38,
      "learning_rate": 4.139561270801816e-05,
      "loss": 0.8446,
      "step": 1822
    },
    {
      "epoch": 1.38,
      "learning_rate": 4.1390885022692896e-05,
      "loss": 0.8825,
      "step": 1823
    },
    {
      "epoch": 1.38,
      "learning_rate": 4.138615733736763e-05,
      "loss": 0.883,
      "step": 1824
    },
    {
      "epoch": 1.38,
      "learning_rate": 4.138142965204237e-05,
      "loss": 0.9703,
      "step": 1825
    },
    {
      "epoch": 1.38,
      "learning_rate": 4.13767019667171e-05,
      "loss": 0.8655,
      "step": 1826
    },
    {
      "epoch": 1.38,
      "learning_rate": 4.137197428139183e-05,
      "loss": 0.8109,
      "step": 1827
    },
    {
      "epoch": 1.38,
      "learning_rate": 4.1367246596066563e-05,
      "loss": 0.7901,
      "step": 1828
    },
    {
      "epoch": 1.38,
      "learning_rate": 4.13625189107413e-05,
      "loss": 0.8347,
      "step": 1829
    },
    {
      "epoch": 1.38,
      "learning_rate": 4.1357791225416034e-05,
      "loss": 0.9142,
      "step": 1830
    },
    {
      "epoch": 1.38,
      "learning_rate": 4.135306354009077e-05,
      "loss": 0.841,
      "step": 1831
    },
    {
      "epoch": 1.39,
      "learning_rate": 4.1348335854765505e-05,
      "loss": 0.9231,
      "step": 1832
    },
    {
      "epoch": 1.39,
      "learning_rate": 4.1343608169440244e-05,
      "loss": 0.8984,
      "step": 1833
    },
    {
      "epoch": 1.39,
      "learning_rate": 4.1338880484114976e-05,
      "loss": 0.9682,
      "step": 1834
    },
    {
      "epoch": 1.39,
      "learning_rate": 4.1334152798789715e-05,
      "loss": 0.7914,
      "step": 1835
    },
    {
      "epoch": 1.39,
      "learning_rate": 4.132942511346445e-05,
      "loss": 0.842,
      "step": 1836
    },
    {
      "epoch": 1.39,
      "learning_rate": 4.1324697428139186e-05,
      "loss": 0.8639,
      "step": 1837
    },
    {
      "epoch": 1.39,
      "learning_rate": 4.131996974281392e-05,
      "loss": 1.0008,
      "step": 1838
    },
    {
      "epoch": 1.39,
      "learning_rate": 4.131524205748866e-05,
      "loss": 0.8679,
      "step": 1839
    },
    {
      "epoch": 1.39,
      "learning_rate": 4.131051437216339e-05,
      "loss": 0.8558,
      "step": 1840
    },
    {
      "epoch": 1.39,
      "learning_rate": 4.130578668683813e-05,
      "loss": 0.8417,
      "step": 1841
    },
    {
      "epoch": 1.39,
      "learning_rate": 4.130105900151286e-05,
      "loss": 0.8867,
      "step": 1842
    },
    {
      "epoch": 1.39,
      "learning_rate": 4.12963313161876e-05,
      "loss": 0.8845,
      "step": 1843
    },
    {
      "epoch": 1.39,
      "learning_rate": 4.129160363086233e-05,
      "loss": 0.872,
      "step": 1844
    },
    {
      "epoch": 1.4,
      "learning_rate": 4.128687594553707e-05,
      "loss": 0.8287,
      "step": 1845
    },
    {
      "epoch": 1.4,
      "learning_rate": 4.12821482602118e-05,
      "loss": 0.8556,
      "step": 1846
    },
    {
      "epoch": 1.4,
      "learning_rate": 4.127742057488654e-05,
      "loss": 0.9239,
      "step": 1847
    },
    {
      "epoch": 1.4,
      "learning_rate": 4.127269288956127e-05,
      "loss": 0.8612,
      "step": 1848
    },
    {
      "epoch": 1.4,
      "learning_rate": 4.126796520423601e-05,
      "loss": 0.8538,
      "step": 1849
    },
    {
      "epoch": 1.4,
      "learning_rate": 4.126323751891074e-05,
      "loss": 0.9497,
      "step": 1850
    },
    {
      "epoch": 1.4,
      "learning_rate": 4.125850983358548e-05,
      "loss": 0.8242,
      "step": 1851
    },
    {
      "epoch": 1.4,
      "learning_rate": 4.1253782148260214e-05,
      "loss": 0.8888,
      "step": 1852
    },
    {
      "epoch": 1.4,
      "learning_rate": 4.1249054462934946e-05,
      "loss": 0.9758,
      "step": 1853
    },
    {
      "epoch": 1.4,
      "learning_rate": 4.1244326777609685e-05,
      "loss": 0.8584,
      "step": 1854
    },
    {
      "epoch": 1.4,
      "learning_rate": 4.123959909228442e-05,
      "loss": 0.8553,
      "step": 1855
    },
    {
      "epoch": 1.4,
      "learning_rate": 4.123487140695915e-05,
      "loss": 0.913,
      "step": 1856
    },
    {
      "epoch": 1.4,
      "learning_rate": 4.123014372163389e-05,
      "loss": 0.9715,
      "step": 1857
    },
    {
      "epoch": 1.41,
      "learning_rate": 4.122541603630862e-05,
      "loss": 0.9554,
      "step": 1858
    },
    {
      "epoch": 1.41,
      "learning_rate": 4.122068835098336e-05,
      "loss": 0.9292,
      "step": 1859
    },
    {
      "epoch": 1.41,
      "learning_rate": 4.121596066565809e-05,
      "loss": 0.8513,
      "step": 1860
    },
    {
      "epoch": 1.41,
      "learning_rate": 4.121123298033283e-05,
      "loss": 0.861,
      "step": 1861
    },
    {
      "epoch": 1.41,
      "learning_rate": 4.120650529500756e-05,
      "loss": 0.8856,
      "step": 1862
    },
    {
      "epoch": 1.41,
      "learning_rate": 4.12017776096823e-05,
      "loss": 0.9274,
      "step": 1863
    },
    {
      "epoch": 1.41,
      "learning_rate": 4.119704992435703e-05,
      "loss": 0.9182,
      "step": 1864
    },
    {
      "epoch": 1.41,
      "learning_rate": 4.119232223903177e-05,
      "loss": 0.8619,
      "step": 1865
    },
    {
      "epoch": 1.41,
      "learning_rate": 4.118759455370651e-05,
      "loss": 0.7559,
      "step": 1866
    },
    {
      "epoch": 1.41,
      "learning_rate": 4.118286686838124e-05,
      "loss": 0.8942,
      "step": 1867
    },
    {
      "epoch": 1.41,
      "learning_rate": 4.117813918305598e-05,
      "loss": 0.9301,
      "step": 1868
    },
    {
      "epoch": 1.41,
      "learning_rate": 4.117341149773071e-05,
      "loss": 0.8775,
      "step": 1869
    },
    {
      "epoch": 1.41,
      "learning_rate": 4.116868381240545e-05,
      "loss": 0.8295,
      "step": 1870
    },
    {
      "epoch": 1.42,
      "learning_rate": 4.1163956127080184e-05,
      "loss": 0.9126,
      "step": 1871
    },
    {
      "epoch": 1.42,
      "learning_rate": 4.115922844175492e-05,
      "loss": 0.9221,
      "step": 1872
    },
    {
      "epoch": 1.42,
      "learning_rate": 4.1154500756429655e-05,
      "loss": 0.9363,
      "step": 1873
    },
    {
      "epoch": 1.42,
      "learning_rate": 4.1149773071104394e-05,
      "loss": 0.8388,
      "step": 1874
    },
    {
      "epoch": 1.42,
      "learning_rate": 4.1145045385779126e-05,
      "loss": 0.8394,
      "step": 1875
    },
    {
      "epoch": 1.42,
      "learning_rate": 4.1140317700453865e-05,
      "loss": 0.801,
      "step": 1876
    },
    {
      "epoch": 1.42,
      "learning_rate": 4.11355900151286e-05,
      "loss": 0.9158,
      "step": 1877
    },
    {
      "epoch": 1.42,
      "learning_rate": 4.1130862329803336e-05,
      "loss": 0.9618,
      "step": 1878
    },
    {
      "epoch": 1.42,
      "learning_rate": 4.112613464447807e-05,
      "loss": 0.8671,
      "step": 1879
    },
    {
      "epoch": 1.42,
      "learning_rate": 4.11214069591528e-05,
      "loss": 0.9781,
      "step": 1880
    },
    {
      "epoch": 1.42,
      "learning_rate": 4.111667927382753e-05,
      "loss": 0.946,
      "step": 1881
    },
    {
      "epoch": 1.42,
      "learning_rate": 4.111195158850227e-05,
      "loss": 1.0592,
      "step": 1882
    },
    {
      "epoch": 1.42,
      "learning_rate": 4.1107223903177e-05,
      "loss": 0.7477,
      "step": 1883
    },
    {
      "epoch": 1.42,
      "learning_rate": 4.110249621785174e-05,
      "loss": 0.8558,
      "step": 1884
    },
    {
      "epoch": 1.43,
      "learning_rate": 4.1097768532526474e-05,
      "loss": 0.866,
      "step": 1885
    },
    {
      "epoch": 1.43,
      "learning_rate": 4.109304084720121e-05,
      "loss": 0.9976,
      "step": 1886
    },
    {
      "epoch": 1.43,
      "learning_rate": 4.1088313161875945e-05,
      "loss": 0.749,
      "step": 1887
    },
    {
      "epoch": 1.43,
      "learning_rate": 4.1083585476550683e-05,
      "loss": 0.8294,
      "step": 1888
    },
    {
      "epoch": 1.43,
      "learning_rate": 4.1078857791225416e-05,
      "loss": 0.85,
      "step": 1889
    },
    {
      "epoch": 1.43,
      "learning_rate": 4.1074130105900154e-05,
      "loss": 0.8707,
      "step": 1890
    },
    {
      "epoch": 1.43,
      "learning_rate": 4.1069402420574886e-05,
      "loss": 0.799,
      "step": 1891
    },
    {
      "epoch": 1.43,
      "learning_rate": 4.1064674735249625e-05,
      "loss": 0.8383,
      "step": 1892
    },
    {
      "epoch": 1.43,
      "learning_rate": 4.105994704992436e-05,
      "loss": 0.9088,
      "step": 1893
    },
    {
      "epoch": 1.43,
      "learning_rate": 4.1055219364599096e-05,
      "loss": 0.8848,
      "step": 1894
    },
    {
      "epoch": 1.43,
      "learning_rate": 4.105049167927383e-05,
      "loss": 0.8608,
      "step": 1895
    },
    {
      "epoch": 1.43,
      "learning_rate": 4.104576399394857e-05,
      "loss": 0.854,
      "step": 1896
    },
    {
      "epoch": 1.43,
      "learning_rate": 4.10410363086233e-05,
      "loss": 0.8757,
      "step": 1897
    },
    {
      "epoch": 1.44,
      "learning_rate": 4.103630862329804e-05,
      "loss": 0.7727,
      "step": 1898
    },
    {
      "epoch": 1.44,
      "learning_rate": 4.103158093797277e-05,
      "loss": 0.7965,
      "step": 1899
    },
    {
      "epoch": 1.44,
      "learning_rate": 4.102685325264751e-05,
      "loss": 0.8643,
      "step": 1900
    },
    {
      "epoch": 1.44,
      "learning_rate": 4.102212556732224e-05,
      "loss": 0.9119,
      "step": 1901
    },
    {
      "epoch": 1.44,
      "learning_rate": 4.101739788199698e-05,
      "loss": 0.9681,
      "step": 1902
    },
    {
      "epoch": 1.44,
      "learning_rate": 4.101267019667171e-05,
      "loss": 0.8877,
      "step": 1903
    },
    {
      "epoch": 1.44,
      "learning_rate": 4.100794251134645e-05,
      "loss": 0.8928,
      "step": 1904
    },
    {
      "epoch": 1.44,
      "learning_rate": 4.100321482602118e-05,
      "loss": 0.937,
      "step": 1905
    },
    {
      "epoch": 1.44,
      "learning_rate": 4.0998487140695915e-05,
      "loss": 0.8523,
      "step": 1906
    },
    {
      "epoch": 1.44,
      "learning_rate": 4.0993759455370654e-05,
      "loss": 0.8269,
      "step": 1907
    },
    {
      "epoch": 1.44,
      "learning_rate": 4.0989031770045386e-05,
      "loss": 0.9149,
      "step": 1908
    },
    {
      "epoch": 1.44,
      "learning_rate": 4.098430408472012e-05,
      "loss": 0.9423,
      "step": 1909
    },
    {
      "epoch": 1.44,
      "learning_rate": 4.0979576399394857e-05,
      "loss": 0.9137,
      "step": 1910
    },
    {
      "epoch": 1.45,
      "learning_rate": 4.097484871406959e-05,
      "loss": 1.0429,
      "step": 1911
    },
    {
      "epoch": 1.45,
      "learning_rate": 4.097012102874433e-05,
      "loss": 0.7896,
      "step": 1912
    },
    {
      "epoch": 1.45,
      "learning_rate": 4.096539334341906e-05,
      "loss": 0.8795,
      "step": 1913
    },
    {
      "epoch": 1.45,
      "learning_rate": 4.09606656580938e-05,
      "loss": 0.7256,
      "step": 1914
    },
    {
      "epoch": 1.45,
      "learning_rate": 4.095593797276853e-05,
      "loss": 0.7359,
      "step": 1915
    },
    {
      "epoch": 1.45,
      "learning_rate": 4.095121028744327e-05,
      "loss": 0.9324,
      "step": 1916
    },
    {
      "epoch": 1.45,
      "learning_rate": 4.0946482602118e-05,
      "loss": 0.8182,
      "step": 1917
    },
    {
      "epoch": 1.45,
      "learning_rate": 4.094175491679274e-05,
      "loss": 0.8664,
      "step": 1918
    },
    {
      "epoch": 1.45,
      "learning_rate": 4.093702723146747e-05,
      "loss": 0.9025,
      "step": 1919
    },
    {
      "epoch": 1.45,
      "learning_rate": 4.093229954614221e-05,
      "loss": 0.7802,
      "step": 1920
    },
    {
      "epoch": 1.45,
      "learning_rate": 4.092757186081694e-05,
      "loss": 0.8502,
      "step": 1921
    },
    {
      "epoch": 1.45,
      "learning_rate": 4.092284417549168e-05,
      "loss": 0.9403,
      "step": 1922
    },
    {
      "epoch": 1.45,
      "learning_rate": 4.0918116490166414e-05,
      "loss": 0.95,
      "step": 1923
    },
    {
      "epoch": 1.46,
      "learning_rate": 4.091338880484115e-05,
      "loss": 0.7992,
      "step": 1924
    },
    {
      "epoch": 1.46,
      "learning_rate": 4.090866111951589e-05,
      "loss": 0.7147,
      "step": 1925
    },
    {
      "epoch": 1.46,
      "learning_rate": 4.0903933434190624e-05,
      "loss": 0.9009,
      "step": 1926
    },
    {
      "epoch": 1.46,
      "learning_rate": 4.089920574886536e-05,
      "loss": 0.8046,
      "step": 1927
    },
    {
      "epoch": 1.46,
      "learning_rate": 4.0894478063540095e-05,
      "loss": 0.8512,
      "step": 1928
    },
    {
      "epoch": 1.46,
      "learning_rate": 4.0889750378214833e-05,
      "loss": 0.8408,
      "step": 1929
    },
    {
      "epoch": 1.46,
      "learning_rate": 4.0885022692889565e-05,
      "loss": 0.8978,
      "step": 1930
    },
    {
      "epoch": 1.46,
      "learning_rate": 4.0880295007564304e-05,
      "loss": 0.7724,
      "step": 1931
    },
    {
      "epoch": 1.46,
      "learning_rate": 4.0875567322239036e-05,
      "loss": 0.8847,
      "step": 1932
    },
    {
      "epoch": 1.46,
      "learning_rate": 4.087083963691377e-05,
      "loss": 0.8122,
      "step": 1933
    },
    {
      "epoch": 1.46,
      "learning_rate": 4.08661119515885e-05,
      "loss": 0.8767,
      "step": 1934
    },
    {
      "epoch": 1.46,
      "learning_rate": 4.086138426626324e-05,
      "loss": 0.752,
      "step": 1935
    },
    {
      "epoch": 1.46,
      "learning_rate": 4.085665658093797e-05,
      "loss": 1.0225,
      "step": 1936
    },
    {
      "epoch": 1.46,
      "learning_rate": 4.085192889561271e-05,
      "loss": 0.8419,
      "step": 1937
    },
    {
      "epoch": 1.47,
      "learning_rate": 4.084720121028744e-05,
      "loss": 0.9642,
      "step": 1938
    },
    {
      "epoch": 1.47,
      "learning_rate": 4.084247352496218e-05,
      "loss": 0.7735,
      "step": 1939
    },
    {
      "epoch": 1.47,
      "learning_rate": 4.083774583963691e-05,
      "loss": 0.876,
      "step": 1940
    },
    {
      "epoch": 1.47,
      "learning_rate": 4.083301815431165e-05,
      "loss": 0.9292,
      "step": 1941
    },
    {
      "epoch": 1.47,
      "learning_rate": 4.0828290468986384e-05,
      "loss": 0.825,
      "step": 1942
    },
    {
      "epoch": 1.47,
      "learning_rate": 4.082356278366112e-05,
      "loss": 0.8153,
      "step": 1943
    },
    {
      "epoch": 1.47,
      "learning_rate": 4.0818835098335855e-05,
      "loss": 0.92,
      "step": 1944
    },
    {
      "epoch": 1.47,
      "learning_rate": 4.0814107413010594e-05,
      "loss": 0.8701,
      "step": 1945
    },
    {
      "epoch": 1.47,
      "learning_rate": 4.0809379727685326e-05,
      "loss": 0.881,
      "step": 1946
    },
    {
      "epoch": 1.47,
      "learning_rate": 4.0804652042360065e-05,
      "loss": 0.9003,
      "step": 1947
    },
    {
      "epoch": 1.47,
      "learning_rate": 4.07999243570348e-05,
      "loss": 0.9212,
      "step": 1948
    },
    {
      "epoch": 1.47,
      "learning_rate": 4.0795196671709536e-05,
      "loss": 0.8815,
      "step": 1949
    },
    {
      "epoch": 1.47,
      "learning_rate": 4.079046898638427e-05,
      "loss": 0.9883,
      "step": 1950
    },
    {
      "epoch": 1.48,
      "learning_rate": 4.0785741301059006e-05,
      "loss": 0.9339,
      "step": 1951
    },
    {
      "epoch": 1.48,
      "learning_rate": 4.078101361573374e-05,
      "loss": 0.8815,
      "step": 1952
    },
    {
      "epoch": 1.48,
      "learning_rate": 4.077628593040848e-05,
      "loss": 0.9142,
      "step": 1953
    },
    {
      "epoch": 1.48,
      "learning_rate": 4.077155824508321e-05,
      "loss": 0.7895,
      "step": 1954
    },
    {
      "epoch": 1.48,
      "learning_rate": 4.076683055975795e-05,
      "loss": 0.8466,
      "step": 1955
    },
    {
      "epoch": 1.48,
      "learning_rate": 4.076210287443268e-05,
      "loss": 0.8297,
      "step": 1956
    },
    {
      "epoch": 1.48,
      "learning_rate": 4.075737518910742e-05,
      "loss": 0.8639,
      "step": 1957
    },
    {
      "epoch": 1.48,
      "learning_rate": 4.075264750378215e-05,
      "loss": 0.8963,
      "step": 1958
    },
    {
      "epoch": 1.48,
      "learning_rate": 4.074791981845688e-05,
      "loss": 0.8327,
      "step": 1959
    },
    {
      "epoch": 1.48,
      "learning_rate": 4.074319213313162e-05,
      "loss": 0.6733,
      "step": 1960
    },
    {
      "epoch": 1.48,
      "learning_rate": 4.0738464447806354e-05,
      "loss": 0.9623,
      "step": 1961
    },
    {
      "epoch": 1.48,
      "learning_rate": 4.0733736762481086e-05,
      "loss": 0.9113,
      "step": 1962
    },
    {
      "epoch": 1.48,
      "learning_rate": 4.0729009077155825e-05,
      "loss": 0.8064,
      "step": 1963
    },
    {
      "epoch": 1.49,
      "learning_rate": 4.072428139183056e-05,
      "loss": 0.8044,
      "step": 1964
    },
    {
      "epoch": 1.49,
      "learning_rate": 4.0719553706505296e-05,
      "loss": 0.7314,
      "step": 1965
    },
    {
      "epoch": 1.49,
      "learning_rate": 4.071482602118003e-05,
      "loss": 0.853,
      "step": 1966
    },
    {
      "epoch": 1.49,
      "learning_rate": 4.071009833585477e-05,
      "loss": 0.8627,
      "step": 1967
    },
    {
      "epoch": 1.49,
      "learning_rate": 4.07053706505295e-05,
      "loss": 0.8834,
      "step": 1968
    },
    {
      "epoch": 1.49,
      "learning_rate": 4.070064296520424e-05,
      "loss": 0.8495,
      "step": 1969
    },
    {
      "epoch": 1.49,
      "learning_rate": 4.069591527987897e-05,
      "loss": 0.8901,
      "step": 1970
    },
    {
      "epoch": 1.49,
      "learning_rate": 4.069118759455371e-05,
      "loss": 0.8028,
      "step": 1971
    },
    {
      "epoch": 1.49,
      "learning_rate": 4.068645990922844e-05,
      "loss": 0.914,
      "step": 1972
    },
    {
      "epoch": 1.49,
      "learning_rate": 4.068173222390318e-05,
      "loss": 0.8204,
      "step": 1973
    },
    {
      "epoch": 1.49,
      "learning_rate": 4.067700453857791e-05,
      "loss": 0.8222,
      "step": 1974
    },
    {
      "epoch": 1.49,
      "learning_rate": 4.067227685325265e-05,
      "loss": 0.9275,
      "step": 1975
    },
    {
      "epoch": 1.49,
      "learning_rate": 4.066754916792738e-05,
      "loss": 0.772,
      "step": 1976
    },
    {
      "epoch": 1.5,
      "learning_rate": 4.066282148260212e-05,
      "loss": 0.9144,
      "step": 1977
    },
    {
      "epoch": 1.5,
      "learning_rate": 4.0658093797276853e-05,
      "loss": 0.8894,
      "step": 1978
    },
    {
      "epoch": 1.5,
      "learning_rate": 4.065336611195159e-05,
      "loss": 0.9927,
      "step": 1979
    },
    {
      "epoch": 1.5,
      "learning_rate": 4.0648638426626324e-05,
      "loss": 0.8588,
      "step": 1980
    },
    {
      "epoch": 1.5,
      "learning_rate": 4.064391074130106e-05,
      "loss": 0.8232,
      "step": 1981
    },
    {
      "epoch": 1.5,
      "learning_rate": 4.0639183055975795e-05,
      "loss": 0.8826,
      "step": 1982
    },
    {
      "epoch": 1.5,
      "learning_rate": 4.0634455370650534e-05,
      "loss": 0.9019,
      "step": 1983
    },
    {
      "epoch": 1.5,
      "learning_rate": 4.0629727685325266e-05,
      "loss": 0.816,
      "step": 1984
    },
    {
      "epoch": 1.5,
      "learning_rate": 4.0625000000000005e-05,
      "loss": 0.9024,
      "step": 1985
    },
    {
      "epoch": 1.5,
      "learning_rate": 4.062027231467474e-05,
      "loss": 0.9572,
      "step": 1986
    },
    {
      "epoch": 1.5,
      "learning_rate": 4.061554462934947e-05,
      "loss": 0.9466,
      "step": 1987
    },
    {
      "epoch": 1.5,
      "learning_rate": 4.061081694402421e-05,
      "loss": 0.9585,
      "step": 1988
    },
    {
      "epoch": 1.5,
      "learning_rate": 4.060608925869894e-05,
      "loss": 0.8959,
      "step": 1989
    },
    {
      "epoch": 1.51,
      "learning_rate": 4.060136157337368e-05,
      "loss": 0.8561,
      "step": 1990
    },
    {
      "epoch": 1.51,
      "learning_rate": 4.059663388804841e-05,
      "loss": 1.0081,
      "step": 1991
    },
    {
      "epoch": 1.51,
      "learning_rate": 4.059190620272315e-05,
      "loss": 0.8836,
      "step": 1992
    },
    {
      "epoch": 1.51,
      "learning_rate": 4.058717851739788e-05,
      "loss": 0.868,
      "step": 1993
    },
    {
      "epoch": 1.51,
      "learning_rate": 4.058245083207262e-05,
      "loss": 0.8922,
      "step": 1994
    },
    {
      "epoch": 1.51,
      "learning_rate": 4.057772314674735e-05,
      "loss": 0.8808,
      "step": 1995
    },
    {
      "epoch": 1.51,
      "learning_rate": 4.057299546142209e-05,
      "loss": 0.8516,
      "step": 1996
    },
    {
      "epoch": 1.51,
      "learning_rate": 4.0568267776096824e-05,
      "loss": 0.9618,
      "step": 1997
    },
    {
      "epoch": 1.51,
      "learning_rate": 4.056354009077156e-05,
      "loss": 0.8685,
      "step": 1998
    },
    {
      "epoch": 1.51,
      "learning_rate": 4.0558812405446294e-05,
      "loss": 0.86,
      "step": 1999
    },
    {
      "epoch": 1.51,
      "learning_rate": 4.055408472012103e-05,
      "loss": 0.8958,
      "step": 2000
    },
    {
      "epoch": 1.51,
      "learning_rate": 4.0549357034795765e-05,
      "loss": 0.7754,
      "step": 2001
    },
    {
      "epoch": 1.51,
      "learning_rate": 4.0544629349470504e-05,
      "loss": 0.8493,
      "step": 2002
    },
    {
      "epoch": 1.51,
      "learning_rate": 4.0539901664145236e-05,
      "loss": 1.0867,
      "step": 2003
    },
    {
      "epoch": 1.52,
      "learning_rate": 4.0535173978819975e-05,
      "loss": 0.8159,
      "step": 2004
    },
    {
      "epoch": 1.52,
      "learning_rate": 4.053044629349471e-05,
      "loss": 0.9025,
      "step": 2005
    },
    {
      "epoch": 1.52,
      "learning_rate": 4.0525718608169446e-05,
      "loss": 0.8709,
      "step": 2006
    },
    {
      "epoch": 1.52,
      "learning_rate": 4.052099092284418e-05,
      "loss": 1.0326,
      "step": 2007
    },
    {
      "epoch": 1.52,
      "learning_rate": 4.051626323751892e-05,
      "loss": 0.888,
      "step": 2008
    },
    {
      "epoch": 1.52,
      "learning_rate": 4.051153555219365e-05,
      "loss": 0.7679,
      "step": 2009
    },
    {
      "epoch": 1.52,
      "learning_rate": 4.050680786686839e-05,
      "loss": 1.0027,
      "step": 2010
    },
    {
      "epoch": 1.52,
      "learning_rate": 4.050208018154312e-05,
      "loss": 0.9306,
      "step": 2011
    },
    {
      "epoch": 1.52,
      "learning_rate": 4.049735249621785e-05,
      "loss": 0.9187,
      "step": 2012
    },
    {
      "epoch": 1.52,
      "learning_rate": 4.049262481089259e-05,
      "loss": 0.8281,
      "step": 2013
    },
    {
      "epoch": 1.52,
      "learning_rate": 4.048789712556732e-05,
      "loss": 0.8903,
      "step": 2014
    },
    {
      "epoch": 1.52,
      "learning_rate": 4.0483169440242055e-05,
      "loss": 0.9241,
      "step": 2015
    },
    {
      "epoch": 1.52,
      "learning_rate": 4.0478441754916794e-05,
      "loss": 0.7207,
      "step": 2016
    },
    {
      "epoch": 1.53,
      "learning_rate": 4.0473714069591526e-05,
      "loss": 0.8928,
      "step": 2017
    },
    {
      "epoch": 1.53,
      "learning_rate": 4.0468986384266265e-05,
      "loss": 0.8489,
      "step": 2018
    },
    {
      "epoch": 1.53,
      "learning_rate": 4.0464258698940997e-05,
      "loss": 0.9054,
      "step": 2019
    },
    {
      "epoch": 1.53,
      "learning_rate": 4.0459531013615735e-05,
      "loss": 0.7678,
      "step": 2020
    },
    {
      "epoch": 1.53,
      "learning_rate": 4.045480332829047e-05,
      "loss": 0.9356,
      "step": 2021
    },
    {
      "epoch": 1.53,
      "learning_rate": 4.0450075642965206e-05,
      "loss": 0.8705,
      "step": 2022
    },
    {
      "epoch": 1.53,
      "learning_rate": 4.044534795763994e-05,
      "loss": 0.8134,
      "step": 2023
    },
    {
      "epoch": 1.53,
      "learning_rate": 4.044062027231468e-05,
      "loss": 0.8827,
      "step": 2024
    },
    {
      "epoch": 1.53,
      "learning_rate": 4.043589258698941e-05,
      "loss": 0.9587,
      "step": 2025
    },
    {
      "epoch": 1.53,
      "learning_rate": 4.043116490166415e-05,
      "loss": 0.8241,
      "step": 2026
    },
    {
      "epoch": 1.53,
      "learning_rate": 4.042643721633888e-05,
      "loss": 0.7205,
      "step": 2027
    },
    {
      "epoch": 1.53,
      "learning_rate": 4.042170953101362e-05,
      "loss": 0.7309,
      "step": 2028
    },
    {
      "epoch": 1.53,
      "learning_rate": 4.041698184568835e-05,
      "loss": 0.889,
      "step": 2029
    },
    {
      "epoch": 1.54,
      "learning_rate": 4.041225416036309e-05,
      "loss": 0.7811,
      "step": 2030
    },
    {
      "epoch": 1.54,
      "learning_rate": 4.040752647503782e-05,
      "loss": 0.9398,
      "step": 2031
    },
    {
      "epoch": 1.54,
      "learning_rate": 4.040279878971256e-05,
      "loss": 0.7926,
      "step": 2032
    },
    {
      "epoch": 1.54,
      "learning_rate": 4.039807110438729e-05,
      "loss": 0.8775,
      "step": 2033
    },
    {
      "epoch": 1.54,
      "learning_rate": 4.039334341906203e-05,
      "loss": 0.8397,
      "step": 2034
    },
    {
      "epoch": 1.54,
      "learning_rate": 4.0388615733736764e-05,
      "loss": 1.0397,
      "step": 2035
    },
    {
      "epoch": 1.54,
      "learning_rate": 4.03838880484115e-05,
      "loss": 0.78,
      "step": 2036
    },
    {
      "epoch": 1.54,
      "learning_rate": 4.0379160363086235e-05,
      "loss": 0.8616,
      "step": 2037
    },
    {
      "epoch": 1.54,
      "learning_rate": 4.0374432677760973e-05,
      "loss": 0.9214,
      "step": 2038
    },
    {
      "epoch": 1.54,
      "learning_rate": 4.0369704992435706e-05,
      "loss": 0.9142,
      "step": 2039
    },
    {
      "epoch": 1.54,
      "learning_rate": 4.036497730711044e-05,
      "loss": 0.7836,
      "step": 2040
    },
    {
      "epoch": 1.54,
      "learning_rate": 4.0360249621785176e-05,
      "loss": 0.8948,
      "step": 2041
    },
    {
      "epoch": 1.54,
      "learning_rate": 4.035552193645991e-05,
      "loss": 0.8195,
      "step": 2042
    },
    {
      "epoch": 1.55,
      "learning_rate": 4.035079425113464e-05,
      "loss": 0.9667,
      "step": 2043
    },
    {
      "epoch": 1.55,
      "learning_rate": 4.034606656580938e-05,
      "loss": 0.8691,
      "step": 2044
    },
    {
      "epoch": 1.55,
      "learning_rate": 4.034133888048411e-05,
      "loss": 0.8193,
      "step": 2045
    },
    {
      "epoch": 1.55,
      "learning_rate": 4.033661119515885e-05,
      "loss": 0.8698,
      "step": 2046
    },
    {
      "epoch": 1.55,
      "learning_rate": 4.033188350983359e-05,
      "loss": 0.8872,
      "step": 2047
    },
    {
      "epoch": 1.55,
      "learning_rate": 4.032715582450832e-05,
      "loss": 0.723,
      "step": 2048
    },
    {
      "epoch": 1.55,
      "learning_rate": 4.032242813918306e-05,
      "loss": 0.8475,
      "step": 2049
    },
    {
      "epoch": 1.55,
      "learning_rate": 4.031770045385779e-05,
      "loss": 0.836,
      "step": 2050
    },
    {
      "epoch": 1.55,
      "learning_rate": 4.031297276853253e-05,
      "loss": 0.9289,
      "step": 2051
    },
    {
      "epoch": 1.55,
      "learning_rate": 4.030824508320726e-05,
      "loss": 0.8425,
      "step": 2052
    },
    {
      "epoch": 1.55,
      "learning_rate": 4.0303517397882e-05,
      "loss": 0.8796,
      "step": 2053
    },
    {
      "epoch": 1.55,
      "learning_rate": 4.0298789712556734e-05,
      "loss": 0.8289,
      "step": 2054
    },
    {
      "epoch": 1.55,
      "learning_rate": 4.029406202723147e-05,
      "loss": 0.9753,
      "step": 2055
    },
    {
      "epoch": 1.55,
      "learning_rate": 4.0289334341906205e-05,
      "loss": 0.8651,
      "step": 2056
    },
    {
      "epoch": 1.56,
      "learning_rate": 4.0284606656580944e-05,
      "loss": 0.7431,
      "step": 2057
    },
    {
      "epoch": 1.56,
      "learning_rate": 4.0279878971255676e-05,
      "loss": 0.7826,
      "step": 2058
    },
    {
      "epoch": 1.56,
      "learning_rate": 4.0275151285930414e-05,
      "loss": 0.9014,
      "step": 2059
    },
    {
      "epoch": 1.56,
      "learning_rate": 4.0270423600605147e-05,
      "loss": 0.8655,
      "step": 2060
    },
    {
      "epoch": 1.56,
      "learning_rate": 4.0265695915279885e-05,
      "loss": 0.8732,
      "step": 2061
    },
    {
      "epoch": 1.56,
      "learning_rate": 4.026096822995462e-05,
      "loss": 0.8887,
      "step": 2062
    },
    {
      "epoch": 1.56,
      "learning_rate": 4.0256240544629356e-05,
      "loss": 0.8817,
      "step": 2063
    },
    {
      "epoch": 1.56,
      "learning_rate": 4.025151285930409e-05,
      "loss": 0.8432,
      "step": 2064
    },
    {
      "epoch": 1.56,
      "learning_rate": 4.024678517397882e-05,
      "loss": 0.9028,
      "step": 2065
    },
    {
      "epoch": 1.56,
      "learning_rate": 4.024205748865356e-05,
      "loss": 0.8547,
      "step": 2066
    },
    {
      "epoch": 1.56,
      "learning_rate": 4.023732980332829e-05,
      "loss": 0.9719,
      "step": 2067
    },
    {
      "epoch": 1.56,
      "learning_rate": 4.023260211800302e-05,
      "loss": 0.9328,
      "step": 2068
    },
    {
      "epoch": 1.56,
      "learning_rate": 4.022787443267776e-05,
      "loss": 0.8597,
      "step": 2069
    },
    {
      "epoch": 1.57,
      "learning_rate": 4.0223146747352494e-05,
      "loss": 0.922,
      "step": 2070
    },
    {
      "epoch": 1.57,
      "learning_rate": 4.021841906202723e-05,
      "loss": 0.9165,
      "step": 2071
    },
    {
      "epoch": 1.57,
      "learning_rate": 4.0213691376701965e-05,
      "loss": 0.8787,
      "step": 2072
    },
    {
      "epoch": 1.57,
      "learning_rate": 4.0208963691376704e-05,
      "loss": 0.8421,
      "step": 2073
    },
    {
      "epoch": 1.57,
      "learning_rate": 4.0204236006051436e-05,
      "loss": 0.9681,
      "step": 2074
    },
    {
      "epoch": 1.57,
      "learning_rate": 4.0199508320726175e-05,
      "loss": 0.8434,
      "step": 2075
    },
    {
      "epoch": 1.57,
      "learning_rate": 4.019478063540091e-05,
      "loss": 0.7603,
      "step": 2076
    },
    {
      "epoch": 1.57,
      "learning_rate": 4.0190052950075646e-05,
      "loss": 0.7594,
      "step": 2077
    },
    {
      "epoch": 1.57,
      "learning_rate": 4.018532526475038e-05,
      "loss": 0.8929,
      "step": 2078
    },
    {
      "epoch": 1.57,
      "learning_rate": 4.018059757942512e-05,
      "loss": 0.832,
      "step": 2079
    },
    {
      "epoch": 1.57,
      "learning_rate": 4.017586989409985e-05,
      "loss": 0.8337,
      "step": 2080
    },
    {
      "epoch": 1.57,
      "learning_rate": 4.017114220877459e-05,
      "loss": 0.848,
      "step": 2081
    },
    {
      "epoch": 1.57,
      "learning_rate": 4.016641452344932e-05,
      "loss": 0.7674,
      "step": 2082
    },
    {
      "epoch": 1.58,
      "learning_rate": 4.016168683812406e-05,
      "loss": 0.8739,
      "step": 2083
    },
    {
      "epoch": 1.58,
      "learning_rate": 4.015695915279879e-05,
      "loss": 0.869,
      "step": 2084
    },
    {
      "epoch": 1.58,
      "learning_rate": 4.015223146747353e-05,
      "loss": 0.8409,
      "step": 2085
    },
    {
      "epoch": 1.58,
      "learning_rate": 4.014750378214826e-05,
      "loss": 0.8541,
      "step": 2086
    },
    {
      "epoch": 1.58,
      "learning_rate": 4.0142776096823e-05,
      "loss": 0.8415,
      "step": 2087
    },
    {
      "epoch": 1.58,
      "learning_rate": 4.013804841149773e-05,
      "loss": 0.9411,
      "step": 2088
    },
    {
      "epoch": 1.58,
      "learning_rate": 4.013332072617247e-05,
      "loss": 0.9232,
      "step": 2089
    },
    {
      "epoch": 1.58,
      "learning_rate": 4.01285930408472e-05,
      "loss": 0.9037,
      "step": 2090
    },
    {
      "epoch": 1.58,
      "learning_rate": 4.012386535552194e-05,
      "loss": 0.8619,
      "step": 2091
    },
    {
      "epoch": 1.58,
      "learning_rate": 4.0119137670196674e-05,
      "loss": 0.7936,
      "step": 2092
    },
    {
      "epoch": 1.58,
      "learning_rate": 4.0114409984871406e-05,
      "loss": 0.9573,
      "step": 2093
    },
    {
      "epoch": 1.58,
      "learning_rate": 4.0109682299546145e-05,
      "loss": 0.8611,
      "step": 2094
    },
    {
      "epoch": 1.58,
      "learning_rate": 4.010495461422088e-05,
      "loss": 0.8914,
      "step": 2095
    },
    {
      "epoch": 1.59,
      "learning_rate": 4.010022692889561e-05,
      "loss": 0.902,
      "step": 2096
    },
    {
      "epoch": 1.59,
      "learning_rate": 4.009549924357035e-05,
      "loss": 0.9043,
      "step": 2097
    },
    {
      "epoch": 1.59,
      "learning_rate": 4.009077155824508e-05,
      "loss": 0.9003,
      "step": 2098
    },
    {
      "epoch": 1.59,
      "learning_rate": 4.008604387291982e-05,
      "loss": 1.1208,
      "step": 2099
    },
    {
      "epoch": 1.59,
      "learning_rate": 4.008131618759455e-05,
      "loss": 0.9452,
      "step": 2100
    },
    {
      "epoch": 1.59,
      "learning_rate": 4.007658850226929e-05,
      "loss": 0.8714,
      "step": 2101
    },
    {
      "epoch": 1.59,
      "learning_rate": 4.007186081694402e-05,
      "loss": 1.0427,
      "step": 2102
    },
    {
      "epoch": 1.59,
      "learning_rate": 4.006713313161876e-05,
      "loss": 0.8495,
      "step": 2103
    },
    {
      "epoch": 1.59,
      "learning_rate": 4.006240544629349e-05,
      "loss": 0.9662,
      "step": 2104
    },
    {
      "epoch": 1.59,
      "learning_rate": 4.005767776096823e-05,
      "loss": 0.8646,
      "step": 2105
    },
    {
      "epoch": 1.59,
      "learning_rate": 4.005295007564297e-05,
      "loss": 0.8651,
      "step": 2106
    },
    {
      "epoch": 1.59,
      "learning_rate": 4.00482223903177e-05,
      "loss": 0.9112,
      "step": 2107
    },
    {
      "epoch": 1.59,
      "learning_rate": 4.004349470499244e-05,
      "loss": 0.878,
      "step": 2108
    },
    {
      "epoch": 1.6,
      "learning_rate": 4.003876701966717e-05,
      "loss": 0.9714,
      "step": 2109
    },
    {
      "epoch": 1.6,
      "learning_rate": 4.003403933434191e-05,
      "loss": 0.9255,
      "step": 2110
    },
    {
      "epoch": 1.6,
      "learning_rate": 4.0029311649016644e-05,
      "loss": 0.9052,
      "step": 2111
    },
    {
      "epoch": 1.6,
      "learning_rate": 4.002458396369138e-05,
      "loss": 0.8659,
      "step": 2112
    },
    {
      "epoch": 1.6,
      "learning_rate": 4.0019856278366115e-05,
      "loss": 0.8954,
      "step": 2113
    },
    {
      "epoch": 1.6,
      "learning_rate": 4.0015128593040854e-05,
      "loss": 0.8912,
      "step": 2114
    },
    {
      "epoch": 1.6,
      "learning_rate": 4.0010400907715586e-05,
      "loss": 0.8858,
      "step": 2115
    },
    {
      "epoch": 1.6,
      "learning_rate": 4.0005673222390325e-05,
      "loss": 0.932,
      "step": 2116
    },
    {
      "epoch": 1.6,
      "learning_rate": 4.000094553706506e-05,
      "loss": 0.8145,
      "step": 2117
    },
    {
      "epoch": 1.6,
      "learning_rate": 3.999621785173979e-05,
      "loss": 0.8868,
      "step": 2118
    },
    {
      "epoch": 1.6,
      "learning_rate": 3.999149016641453e-05,
      "loss": 0.9773,
      "step": 2119
    },
    {
      "epoch": 1.6,
      "learning_rate": 3.998676248108926e-05,
      "loss": 0.9682,
      "step": 2120
    },
    {
      "epoch": 1.6,
      "learning_rate": 3.998203479576399e-05,
      "loss": 0.9856,
      "step": 2121
    },
    {
      "epoch": 1.6,
      "learning_rate": 3.997730711043873e-05,
      "loss": 0.8498,
      "step": 2122
    },
    {
      "epoch": 1.61,
      "learning_rate": 3.997257942511346e-05,
      "loss": 0.7674,
      "step": 2123
    },
    {
      "epoch": 1.61,
      "learning_rate": 3.99678517397882e-05,
      "loss": 0.9605,
      "step": 2124
    },
    {
      "epoch": 1.61,
      "learning_rate": 3.9963124054462934e-05,
      "loss": 0.9943,
      "step": 2125
    },
    {
      "epoch": 1.61,
      "learning_rate": 3.995839636913767e-05,
      "loss": 0.7984,
      "step": 2126
    },
    {
      "epoch": 1.61,
      "learning_rate": 3.9953668683812405e-05,
      "loss": 0.7934,
      "step": 2127
    },
    {
      "epoch": 1.61,
      "learning_rate": 3.9948940998487143e-05,
      "loss": 0.8784,
      "step": 2128
    },
    {
      "epoch": 1.61,
      "learning_rate": 3.9944213313161875e-05,
      "loss": 0.8796,
      "step": 2129
    },
    {
      "epoch": 1.61,
      "learning_rate": 3.9939485627836614e-05,
      "loss": 0.7721,
      "step": 2130
    },
    {
      "epoch": 1.61,
      "learning_rate": 3.9934757942511346e-05,
      "loss": 0.9686,
      "step": 2131
    },
    {
      "epoch": 1.61,
      "learning_rate": 3.9930030257186085e-05,
      "loss": 0.8049,
      "step": 2132
    },
    {
      "epoch": 1.61,
      "learning_rate": 3.992530257186082e-05,
      "loss": 0.8209,
      "step": 2133
    },
    {
      "epoch": 1.61,
      "learning_rate": 3.9920574886535556e-05,
      "loss": 0.7988,
      "step": 2134
    },
    {
      "epoch": 1.61,
      "learning_rate": 3.991584720121029e-05,
      "loss": 1.0067,
      "step": 2135
    },
    {
      "epoch": 1.62,
      "learning_rate": 3.991111951588503e-05,
      "loss": 0.9827,
      "step": 2136
    },
    {
      "epoch": 1.62,
      "learning_rate": 3.990639183055976e-05,
      "loss": 0.839,
      "step": 2137
    },
    {
      "epoch": 1.62,
      "learning_rate": 3.99016641452345e-05,
      "loss": 0.9259,
      "step": 2138
    },
    {
      "epoch": 1.62,
      "learning_rate": 3.989693645990923e-05,
      "loss": 0.7611,
      "step": 2139
    },
    {
      "epoch": 1.62,
      "learning_rate": 3.989220877458397e-05,
      "loss": 0.7979,
      "step": 2140
    },
    {
      "epoch": 1.62,
      "learning_rate": 3.98874810892587e-05,
      "loss": 0.7124,
      "step": 2141
    },
    {
      "epoch": 1.62,
      "learning_rate": 3.988275340393344e-05,
      "loss": 0.7372,
      "step": 2142
    },
    {
      "epoch": 1.62,
      "learning_rate": 3.987802571860817e-05,
      "loss": 0.9213,
      "step": 2143
    },
    {
      "epoch": 1.62,
      "learning_rate": 3.987329803328291e-05,
      "loss": 0.8475,
      "step": 2144
    },
    {
      "epoch": 1.62,
      "learning_rate": 3.986857034795764e-05,
      "loss": 0.9251,
      "step": 2145
    },
    {
      "epoch": 1.62,
      "learning_rate": 3.9863842662632375e-05,
      "loss": 0.9744,
      "step": 2146
    },
    {
      "epoch": 1.62,
      "learning_rate": 3.9859114977307114e-05,
      "loss": 0.9254,
      "step": 2147
    },
    {
      "epoch": 1.62,
      "learning_rate": 3.9854387291981846e-05,
      "loss": 0.8635,
      "step": 2148
    },
    {
      "epoch": 1.63,
      "learning_rate": 3.984965960665658e-05,
      "loss": 0.9679,
      "step": 2149
    },
    {
      "epoch": 1.63,
      "learning_rate": 3.9844931921331316e-05,
      "loss": 0.9023,
      "step": 2150
    },
    {
      "epoch": 1.63,
      "learning_rate": 3.984020423600605e-05,
      "loss": 1.0302,
      "step": 2151
    },
    {
      "epoch": 1.63,
      "learning_rate": 3.983547655068079e-05,
      "loss": 0.9645,
      "step": 2152
    },
    {
      "epoch": 1.63,
      "learning_rate": 3.983074886535552e-05,
      "loss": 0.8523,
      "step": 2153
    },
    {
      "epoch": 1.63,
      "learning_rate": 3.982602118003026e-05,
      "loss": 0.8524,
      "step": 2154
    },
    {
      "epoch": 1.63,
      "learning_rate": 3.982129349470499e-05,
      "loss": 0.8858,
      "step": 2155
    },
    {
      "epoch": 1.63,
      "learning_rate": 3.981656580937973e-05,
      "loss": 0.8843,
      "step": 2156
    },
    {
      "epoch": 1.63,
      "learning_rate": 3.981183812405446e-05,
      "loss": 0.8974,
      "step": 2157
    },
    {
      "epoch": 1.63,
      "learning_rate": 3.98071104387292e-05,
      "loss": 0.9414,
      "step": 2158
    },
    {
      "epoch": 1.63,
      "learning_rate": 3.980238275340393e-05,
      "loss": 0.7591,
      "step": 2159
    },
    {
      "epoch": 1.63,
      "learning_rate": 3.979765506807867e-05,
      "loss": 0.9019,
      "step": 2160
    },
    {
      "epoch": 1.63,
      "learning_rate": 3.97929273827534e-05,
      "loss": 0.791,
      "step": 2161
    },
    {
      "epoch": 1.64,
      "learning_rate": 3.978819969742814e-05,
      "loss": 0.9155,
      "step": 2162
    },
    {
      "epoch": 1.64,
      "learning_rate": 3.9783472012102874e-05,
      "loss": 0.7904,
      "step": 2163
    },
    {
      "epoch": 1.64,
      "learning_rate": 3.977874432677761e-05,
      "loss": 0.8167,
      "step": 2164
    },
    {
      "epoch": 1.64,
      "learning_rate": 3.9774016641452345e-05,
      "loss": 0.8157,
      "step": 2165
    },
    {
      "epoch": 1.64,
      "learning_rate": 3.9769288956127084e-05,
      "loss": 1.0124,
      "step": 2166
    },
    {
      "epoch": 1.64,
      "learning_rate": 3.976456127080182e-05,
      "loss": 0.7555,
      "step": 2167
    },
    {
      "epoch": 1.64,
      "learning_rate": 3.9759833585476555e-05,
      "loss": 0.802,
      "step": 2168
    },
    {
      "epoch": 1.64,
      "learning_rate": 3.975510590015129e-05,
      "loss": 0.8698,
      "step": 2169
    },
    {
      "epoch": 1.64,
      "learning_rate": 3.9750378214826025e-05,
      "loss": 0.853,
      "step": 2170
    },
    {
      "epoch": 1.64,
      "learning_rate": 3.974565052950076e-05,
      "loss": 0.867,
      "step": 2171
    },
    {
      "epoch": 1.64,
      "learning_rate": 3.9740922844175496e-05,
      "loss": 0.9911,
      "step": 2172
    },
    {
      "epoch": 1.64,
      "learning_rate": 3.973619515885023e-05,
      "loss": 0.8023,
      "step": 2173
    },
    {
      "epoch": 1.64,
      "learning_rate": 3.973146747352496e-05,
      "loss": 0.957,
      "step": 2174
    },
    {
      "epoch": 1.64,
      "learning_rate": 3.97267397881997e-05,
      "loss": 0.9008,
      "step": 2175
    },
    {
      "epoch": 1.65,
      "learning_rate": 3.972201210287443e-05,
      "loss": 0.9984,
      "step": 2176
    },
    {
      "epoch": 1.65,
      "learning_rate": 3.971728441754917e-05,
      "loss": 0.7942,
      "step": 2177
    },
    {
      "epoch": 1.65,
      "learning_rate": 3.97125567322239e-05,
      "loss": 0.7996,
      "step": 2178
    },
    {
      "epoch": 1.65,
      "learning_rate": 3.970782904689864e-05,
      "loss": 0.8009,
      "step": 2179
    },
    {
      "epoch": 1.65,
      "learning_rate": 3.970310136157337e-05,
      "loss": 0.9026,
      "step": 2180
    },
    {
      "epoch": 1.65,
      "learning_rate": 3.969837367624811e-05,
      "loss": 0.8435,
      "step": 2181
    },
    {
      "epoch": 1.65,
      "learning_rate": 3.9693645990922844e-05,
      "loss": 1.113,
      "step": 2182
    },
    {
      "epoch": 1.65,
      "learning_rate": 3.968891830559758e-05,
      "loss": 0.7831,
      "step": 2183
    },
    {
      "epoch": 1.65,
      "learning_rate": 3.9684190620272315e-05,
      "loss": 0.8861,
      "step": 2184
    },
    {
      "epoch": 1.65,
      "learning_rate": 3.9679462934947054e-05,
      "loss": 0.9109,
      "step": 2185
    },
    {
      "epoch": 1.65,
      "learning_rate": 3.9674735249621786e-05,
      "loss": 0.9598,
      "step": 2186
    },
    {
      "epoch": 1.65,
      "learning_rate": 3.9670007564296525e-05,
      "loss": 0.9367,
      "step": 2187
    },
    {
      "epoch": 1.65,
      "learning_rate": 3.966527987897126e-05,
      "loss": 0.8579,
      "step": 2188
    },
    {
      "epoch": 1.66,
      "learning_rate": 3.9660552193645996e-05,
      "loss": 0.938,
      "step": 2189
    },
    {
      "epoch": 1.66,
      "learning_rate": 3.965582450832073e-05,
      "loss": 0.8299,
      "step": 2190
    },
    {
      "epoch": 1.66,
      "learning_rate": 3.9651096822995466e-05,
      "loss": 1.0076,
      "step": 2191
    },
    {
      "epoch": 1.66,
      "learning_rate": 3.96463691376702e-05,
      "loss": 0.9475,
      "step": 2192
    },
    {
      "epoch": 1.66,
      "learning_rate": 3.964164145234494e-05,
      "loss": 0.9665,
      "step": 2193
    },
    {
      "epoch": 1.66,
      "learning_rate": 3.963691376701967e-05,
      "loss": 0.8982,
      "step": 2194
    },
    {
      "epoch": 1.66,
      "learning_rate": 3.963218608169441e-05,
      "loss": 0.8808,
      "step": 2195
    },
    {
      "epoch": 1.66,
      "learning_rate": 3.962745839636914e-05,
      "loss": 0.887,
      "step": 2196
    },
    {
      "epoch": 1.66,
      "learning_rate": 3.962273071104388e-05,
      "loss": 0.835,
      "step": 2197
    },
    {
      "epoch": 1.66,
      "learning_rate": 3.961800302571861e-05,
      "loss": 1.0036,
      "step": 2198
    },
    {
      "epoch": 1.66,
      "learning_rate": 3.961327534039334e-05,
      "loss": 0.9091,
      "step": 2199
    },
    {
      "epoch": 1.66,
      "learning_rate": 3.960854765506808e-05,
      "loss": 0.9475,
      "step": 2200
    },
    {
      "epoch": 1.66,
      "learning_rate": 3.9603819969742814e-05,
      "loss": 0.8709,
      "step": 2201
    },
    {
      "epoch": 1.67,
      "learning_rate": 3.9599092284417546e-05,
      "loss": 0.9693,
      "step": 2202
    },
    {
      "epoch": 1.67,
      "learning_rate": 3.9594364599092285e-05,
      "loss": 0.9573,
      "step": 2203
    },
    {
      "epoch": 1.67,
      "learning_rate": 3.958963691376702e-05,
      "loss": 0.8179,
      "step": 2204
    },
    {
      "epoch": 1.67,
      "learning_rate": 3.9584909228441756e-05,
      "loss": 0.8547,
      "step": 2205
    },
    {
      "epoch": 1.67,
      "learning_rate": 3.958018154311649e-05,
      "loss": 0.9651,
      "step": 2206
    },
    {
      "epoch": 1.67,
      "learning_rate": 3.957545385779123e-05,
      "loss": 0.8468,
      "step": 2207
    },
    {
      "epoch": 1.67,
      "learning_rate": 3.957072617246596e-05,
      "loss": 0.7702,
      "step": 2208
    },
    {
      "epoch": 1.67,
      "learning_rate": 3.95659984871407e-05,
      "loss": 0.9069,
      "step": 2209
    },
    {
      "epoch": 1.67,
      "learning_rate": 3.956127080181543e-05,
      "loss": 0.969,
      "step": 2210
    },
    {
      "epoch": 1.67,
      "learning_rate": 3.955654311649017e-05,
      "loss": 0.9476,
      "step": 2211
    },
    {
      "epoch": 1.67,
      "learning_rate": 3.95518154311649e-05,
      "loss": 1.0313,
      "step": 2212
    },
    {
      "epoch": 1.67,
      "learning_rate": 3.954708774583964e-05,
      "loss": 0.9287,
      "step": 2213
    },
    {
      "epoch": 1.67,
      "learning_rate": 3.954236006051437e-05,
      "loss": 0.8581,
      "step": 2214
    },
    {
      "epoch": 1.68,
      "learning_rate": 3.953763237518911e-05,
      "loss": 0.8319,
      "step": 2215
    },
    {
      "epoch": 1.68,
      "learning_rate": 3.953290468986384e-05,
      "loss": 0.9595,
      "step": 2216
    },
    {
      "epoch": 1.68,
      "learning_rate": 3.952817700453858e-05,
      "loss": 0.8777,
      "step": 2217
    },
    {
      "epoch": 1.68,
      "learning_rate": 3.952344931921331e-05,
      "loss": 0.824,
      "step": 2218
    },
    {
      "epoch": 1.68,
      "learning_rate": 3.951872163388805e-05,
      "loss": 0.8014,
      "step": 2219
    },
    {
      "epoch": 1.68,
      "learning_rate": 3.9513993948562784e-05,
      "loss": 0.8735,
      "step": 2220
    },
    {
      "epoch": 1.68,
      "learning_rate": 3.950926626323752e-05,
      "loss": 0.9224,
      "step": 2221
    },
    {
      "epoch": 1.68,
      "learning_rate": 3.9504538577912255e-05,
      "loss": 0.8889,
      "step": 2222
    },
    {
      "epoch": 1.68,
      "learning_rate": 3.9499810892586994e-05,
      "loss": 0.871,
      "step": 2223
    },
    {
      "epoch": 1.68,
      "learning_rate": 3.9495083207261726e-05,
      "loss": 0.849,
      "step": 2224
    },
    {
      "epoch": 1.68,
      "learning_rate": 3.9490355521936465e-05,
      "loss": 0.8659,
      "step": 2225
    },
    {
      "epoch": 1.68,
      "learning_rate": 3.94856278366112e-05,
      "loss": 0.9019,
      "step": 2226
    },
    {
      "epoch": 1.68,
      "learning_rate": 3.948090015128593e-05,
      "loss": 0.8529,
      "step": 2227
    },
    {
      "epoch": 1.69,
      "learning_rate": 3.947617246596067e-05,
      "loss": 0.832,
      "step": 2228
    },
    {
      "epoch": 1.69,
      "learning_rate": 3.94714447806354e-05,
      "loss": 0.9185,
      "step": 2229
    },
    {
      "epoch": 1.69,
      "learning_rate": 3.946671709531014e-05,
      "loss": 0.9047,
      "step": 2230
    },
    {
      "epoch": 1.69,
      "learning_rate": 3.946198940998487e-05,
      "loss": 0.8973,
      "step": 2231
    },
    {
      "epoch": 1.69,
      "learning_rate": 3.945726172465961e-05,
      "loss": 0.835,
      "step": 2232
    },
    {
      "epoch": 1.69,
      "learning_rate": 3.945253403933434e-05,
      "loss": 0.8896,
      "step": 2233
    },
    {
      "epoch": 1.69,
      "learning_rate": 3.944780635400908e-05,
      "loss": 0.7897,
      "step": 2234
    },
    {
      "epoch": 1.69,
      "learning_rate": 3.944307866868381e-05,
      "loss": 0.9194,
      "step": 2235
    },
    {
      "epoch": 1.69,
      "learning_rate": 3.943835098335855e-05,
      "loss": 0.9242,
      "step": 2236
    },
    {
      "epoch": 1.69,
      "learning_rate": 3.9433623298033283e-05,
      "loss": 0.8781,
      "step": 2237
    },
    {
      "epoch": 1.69,
      "learning_rate": 3.942889561270802e-05,
      "loss": 0.8574,
      "step": 2238
    },
    {
      "epoch": 1.69,
      "learning_rate": 3.9424167927382754e-05,
      "loss": 1.0155,
      "step": 2239
    },
    {
      "epoch": 1.69,
      "learning_rate": 3.941944024205749e-05,
      "loss": 1.0315,
      "step": 2240
    },
    {
      "epoch": 1.69,
      "learning_rate": 3.9414712556732225e-05,
      "loss": 0.8786,
      "step": 2241
    },
    {
      "epoch": 1.7,
      "learning_rate": 3.9409984871406964e-05,
      "loss": 0.927,
      "step": 2242
    },
    {
      "epoch": 1.7,
      "learning_rate": 3.9405257186081696e-05,
      "loss": 0.9213,
      "step": 2243
    },
    {
      "epoch": 1.7,
      "learning_rate": 3.9400529500756435e-05,
      "loss": 0.9205,
      "step": 2244
    },
    {
      "epoch": 1.7,
      "learning_rate": 3.939580181543117e-05,
      "loss": 0.8692,
      "step": 2245
    },
    {
      "epoch": 1.7,
      "learning_rate": 3.9391074130105906e-05,
      "loss": 0.8017,
      "step": 2246
    },
    {
      "epoch": 1.7,
      "learning_rate": 3.938634644478064e-05,
      "loss": 0.9434,
      "step": 2247
    },
    {
      "epoch": 1.7,
      "learning_rate": 3.938161875945538e-05,
      "loss": 0.9414,
      "step": 2248
    },
    {
      "epoch": 1.7,
      "learning_rate": 3.937689107413011e-05,
      "loss": 0.9016,
      "step": 2249
    },
    {
      "epoch": 1.7,
      "learning_rate": 3.937216338880485e-05,
      "loss": 0.9244,
      "step": 2250
    },
    {
      "epoch": 1.7,
      "learning_rate": 3.936743570347958e-05,
      "loss": 0.8794,
      "step": 2251
    },
    {
      "epoch": 1.7,
      "learning_rate": 3.936270801815431e-05,
      "loss": 0.9421,
      "step": 2252
    },
    {
      "epoch": 1.7,
      "learning_rate": 3.935798033282905e-05,
      "loss": 0.8737,
      "step": 2253
    },
    {
      "epoch": 1.7,
      "learning_rate": 3.935325264750378e-05,
      "loss": 0.882,
      "step": 2254
    },
    {
      "epoch": 1.71,
      "learning_rate": 3.9348524962178515e-05,
      "loss": 0.813,
      "step": 2255
    },
    {
      "epoch": 1.71,
      "learning_rate": 3.9343797276853254e-05,
      "loss": 0.9108,
      "step": 2256
    },
    {
      "epoch": 1.71,
      "learning_rate": 3.9339069591527986e-05,
      "loss": 0.9284,
      "step": 2257
    },
    {
      "epoch": 1.71,
      "learning_rate": 3.9334341906202724e-05,
      "loss": 0.8494,
      "step": 2258
    },
    {
      "epoch": 1.71,
      "learning_rate": 3.9329614220877457e-05,
      "loss": 0.8387,
      "step": 2259
    },
    {
      "epoch": 1.71,
      "learning_rate": 3.9324886535552195e-05,
      "loss": 0.844,
      "step": 2260
    },
    {
      "epoch": 1.71,
      "learning_rate": 3.932015885022693e-05,
      "loss": 0.9059,
      "step": 2261
    },
    {
      "epoch": 1.71,
      "learning_rate": 3.9315431164901666e-05,
      "loss": 0.8379,
      "step": 2262
    },
    {
      "epoch": 1.71,
      "learning_rate": 3.93107034795764e-05,
      "loss": 0.9846,
      "step": 2263
    },
    {
      "epoch": 1.71,
      "learning_rate": 3.930597579425114e-05,
      "loss": 0.9601,
      "step": 2264
    },
    {
      "epoch": 1.71,
      "learning_rate": 3.930124810892587e-05,
      "loss": 0.9223,
      "step": 2265
    },
    {
      "epoch": 1.71,
      "learning_rate": 3.929652042360061e-05,
      "loss": 0.8725,
      "step": 2266
    },
    {
      "epoch": 1.71,
      "learning_rate": 3.929179273827534e-05,
      "loss": 0.8749,
      "step": 2267
    },
    {
      "epoch": 1.72,
      "learning_rate": 3.928706505295008e-05,
      "loss": 0.8076,
      "step": 2268
    },
    {
      "epoch": 1.72,
      "learning_rate": 3.928233736762481e-05,
      "loss": 1.0135,
      "step": 2269
    },
    {
      "epoch": 1.72,
      "learning_rate": 3.927760968229955e-05,
      "loss": 0.9563,
      "step": 2270
    },
    {
      "epoch": 1.72,
      "learning_rate": 3.927288199697428e-05,
      "loss": 1.0128,
      "step": 2271
    },
    {
      "epoch": 1.72,
      "learning_rate": 3.926815431164902e-05,
      "loss": 0.9578,
      "step": 2272
    },
    {
      "epoch": 1.72,
      "learning_rate": 3.926342662632375e-05,
      "loss": 0.9611,
      "step": 2273
    },
    {
      "epoch": 1.72,
      "learning_rate": 3.925869894099849e-05,
      "loss": 0.8329,
      "step": 2274
    },
    {
      "epoch": 1.72,
      "learning_rate": 3.9253971255673224e-05,
      "loss": 0.8855,
      "step": 2275
    },
    {
      "epoch": 1.72,
      "learning_rate": 3.924924357034796e-05,
      "loss": 0.8165,
      "step": 2276
    },
    {
      "epoch": 1.72,
      "learning_rate": 3.9244515885022695e-05,
      "loss": 0.8454,
      "step": 2277
    },
    {
      "epoch": 1.72,
      "learning_rate": 3.9239788199697433e-05,
      "loss": 0.8214,
      "step": 2278
    },
    {
      "epoch": 1.72,
      "learning_rate": 3.9235060514372165e-05,
      "loss": 0.896,
      "step": 2279
    },
    {
      "epoch": 1.72,
      "learning_rate": 3.92303328290469e-05,
      "loss": 0.8224,
      "step": 2280
    },
    {
      "epoch": 1.73,
      "learning_rate": 3.9225605143721636e-05,
      "loss": 0.9742,
      "step": 2281
    },
    {
      "epoch": 1.73,
      "learning_rate": 3.922087745839637e-05,
      "loss": 0.8543,
      "step": 2282
    },
    {
      "epoch": 1.73,
      "learning_rate": 3.92161497730711e-05,
      "loss": 0.9206,
      "step": 2283
    },
    {
      "epoch": 1.73,
      "learning_rate": 3.921142208774584e-05,
      "loss": 0.8165,
      "step": 2284
    },
    {
      "epoch": 1.73,
      "learning_rate": 3.920669440242057e-05,
      "loss": 0.838,
      "step": 2285
    },
    {
      "epoch": 1.73,
      "learning_rate": 3.920196671709531e-05,
      "loss": 0.9124,
      "step": 2286
    },
    {
      "epoch": 1.73,
      "learning_rate": 3.919723903177005e-05,
      "loss": 0.9111,
      "step": 2287
    },
    {
      "epoch": 1.73,
      "learning_rate": 3.919251134644478e-05,
      "loss": 0.7956,
      "step": 2288
    },
    {
      "epoch": 1.73,
      "learning_rate": 3.918778366111952e-05,
      "loss": 1.0138,
      "step": 2289
    },
    {
      "epoch": 1.73,
      "learning_rate": 3.918305597579425e-05,
      "loss": 0.9031,
      "step": 2290
    },
    {
      "epoch": 1.73,
      "learning_rate": 3.917832829046899e-05,
      "loss": 0.7912,
      "step": 2291
    },
    {
      "epoch": 1.73,
      "learning_rate": 3.917360060514372e-05,
      "loss": 0.8956,
      "step": 2292
    },
    {
      "epoch": 1.73,
      "learning_rate": 3.916887291981846e-05,
      "loss": 0.9126,
      "step": 2293
    },
    {
      "epoch": 1.73,
      "learning_rate": 3.9164145234493194e-05,
      "loss": 0.8896,
      "step": 2294
    },
    {
      "epoch": 1.74,
      "learning_rate": 3.915941754916793e-05,
      "loss": 0.845,
      "step": 2295
    },
    {
      "epoch": 1.74,
      "learning_rate": 3.9154689863842665e-05,
      "loss": 0.8711,
      "step": 2296
    },
    {
      "epoch": 1.74,
      "learning_rate": 3.9149962178517404e-05,
      "loss": 0.8625,
      "step": 2297
    },
    {
      "epoch": 1.74,
      "learning_rate": 3.9145234493192136e-05,
      "loss": 0.894,
      "step": 2298
    },
    {
      "epoch": 1.74,
      "learning_rate": 3.9140506807866874e-05,
      "loss": 0.8482,
      "step": 2299
    },
    {
      "epoch": 1.74,
      "learning_rate": 3.9135779122541606e-05,
      "loss": 0.9069,
      "step": 2300
    },
    {
      "epoch": 1.74,
      "learning_rate": 3.9131051437216345e-05,
      "loss": 0.8675,
      "step": 2301
    },
    {
      "epoch": 1.74,
      "learning_rate": 3.912632375189108e-05,
      "loss": 0.8388,
      "step": 2302
    },
    {
      "epoch": 1.74,
      "learning_rate": 3.9121596066565816e-05,
      "loss": 0.8515,
      "step": 2303
    },
    {
      "epoch": 1.74,
      "learning_rate": 3.911686838124055e-05,
      "loss": 0.9053,
      "step": 2304
    },
    {
      "epoch": 1.74,
      "learning_rate": 3.911214069591528e-05,
      "loss": 0.886,
      "step": 2305
    },
    {
      "epoch": 1.74,
      "learning_rate": 3.910741301059002e-05,
      "loss": 0.855,
      "step": 2306
    },
    {
      "epoch": 1.74,
      "learning_rate": 3.910268532526475e-05,
      "loss": 0.9229,
      "step": 2307
    },
    {
      "epoch": 1.75,
      "learning_rate": 3.909795763993948e-05,
      "loss": 0.8656,
      "step": 2308
    },
    {
      "epoch": 1.75,
      "learning_rate": 3.909322995461422e-05,
      "loss": 0.895,
      "step": 2309
    },
    {
      "epoch": 1.75,
      "learning_rate": 3.9088502269288954e-05,
      "loss": 0.7788,
      "step": 2310
    },
    {
      "epoch": 1.75,
      "learning_rate": 3.908377458396369e-05,
      "loss": 1.0032,
      "step": 2311
    },
    {
      "epoch": 1.75,
      "learning_rate": 3.9079046898638425e-05,
      "loss": 0.9324,
      "step": 2312
    },
    {
      "epoch": 1.75,
      "learning_rate": 3.9074319213313164e-05,
      "loss": 0.8226,
      "step": 2313
    },
    {
      "epoch": 1.75,
      "learning_rate": 3.9069591527987896e-05,
      "loss": 0.9094,
      "step": 2314
    },
    {
      "epoch": 1.75,
      "learning_rate": 3.9064863842662635e-05,
      "loss": 0.7102,
      "step": 2315
    },
    {
      "epoch": 1.75,
      "learning_rate": 3.906013615733737e-05,
      "loss": 0.9703,
      "step": 2316
    },
    {
      "epoch": 1.75,
      "learning_rate": 3.9055408472012106e-05,
      "loss": 0.9486,
      "step": 2317
    },
    {
      "epoch": 1.75,
      "learning_rate": 3.905068078668684e-05,
      "loss": 0.9755,
      "step": 2318
    },
    {
      "epoch": 1.75,
      "learning_rate": 3.9045953101361577e-05,
      "loss": 0.9066,
      "step": 2319
    },
    {
      "epoch": 1.75,
      "learning_rate": 3.904122541603631e-05,
      "loss": 0.9264,
      "step": 2320
    },
    {
      "epoch": 1.76,
      "learning_rate": 3.903649773071105e-05,
      "loss": 0.801,
      "step": 2321
    },
    {
      "epoch": 1.76,
      "learning_rate": 3.903177004538578e-05,
      "loss": 0.8293,
      "step": 2322
    },
    {
      "epoch": 1.76,
      "learning_rate": 3.902704236006052e-05,
      "loss": 0.9073,
      "step": 2323
    },
    {
      "epoch": 1.76,
      "learning_rate": 3.902231467473525e-05,
      "loss": 0.8397,
      "step": 2324
    },
    {
      "epoch": 1.76,
      "learning_rate": 3.901758698940999e-05,
      "loss": 0.715,
      "step": 2325
    },
    {
      "epoch": 1.76,
      "learning_rate": 3.901285930408472e-05,
      "loss": 0.9946,
      "step": 2326
    },
    {
      "epoch": 1.76,
      "learning_rate": 3.900813161875946e-05,
      "loss": 0.8213,
      "step": 2327
    },
    {
      "epoch": 1.76,
      "learning_rate": 3.900340393343419e-05,
      "loss": 0.8184,
      "step": 2328
    },
    {
      "epoch": 1.76,
      "learning_rate": 3.899867624810893e-05,
      "loss": 0.839,
      "step": 2329
    },
    {
      "epoch": 1.76,
      "learning_rate": 3.899394856278366e-05,
      "loss": 0.8934,
      "step": 2330
    },
    {
      "epoch": 1.76,
      "learning_rate": 3.89892208774584e-05,
      "loss": 0.8306,
      "step": 2331
    },
    {
      "epoch": 1.76,
      "learning_rate": 3.8984493192133134e-05,
      "loss": 0.8846,
      "step": 2332
    },
    {
      "epoch": 1.76,
      "learning_rate": 3.8979765506807866e-05,
      "loss": 0.844,
      "step": 2333
    },
    {
      "epoch": 1.77,
      "learning_rate": 3.8975037821482605e-05,
      "loss": 0.9179,
      "step": 2334
    },
    {
      "epoch": 1.77,
      "learning_rate": 3.897031013615734e-05,
      "loss": 0.9633,
      "step": 2335
    },
    {
      "epoch": 1.77,
      "learning_rate": 3.896558245083207e-05,
      "loss": 0.8701,
      "step": 2336
    },
    {
      "epoch": 1.77,
      "learning_rate": 3.896085476550681e-05,
      "loss": 0.9482,
      "step": 2337
    },
    {
      "epoch": 1.77,
      "learning_rate": 3.895612708018154e-05,
      "loss": 0.8473,
      "step": 2338
    },
    {
      "epoch": 1.77,
      "learning_rate": 3.895139939485628e-05,
      "loss": 0.9663,
      "step": 2339
    },
    {
      "epoch": 1.77,
      "learning_rate": 3.894667170953101e-05,
      "loss": 0.7868,
      "step": 2340
    },
    {
      "epoch": 1.77,
      "learning_rate": 3.894194402420575e-05,
      "loss": 0.9282,
      "step": 2341
    },
    {
      "epoch": 1.77,
      "learning_rate": 3.893721633888048e-05,
      "loss": 0.9044,
      "step": 2342
    },
    {
      "epoch": 1.77,
      "learning_rate": 3.893248865355522e-05,
      "loss": 0.9798,
      "step": 2343
    },
    {
      "epoch": 1.77,
      "learning_rate": 3.892776096822995e-05,
      "loss": 0.9039,
      "step": 2344
    },
    {
      "epoch": 1.77,
      "learning_rate": 3.892303328290469e-05,
      "loss": 0.8795,
      "step": 2345
    },
    {
      "epoch": 1.77,
      "learning_rate": 3.891830559757943e-05,
      "loss": 0.9601,
      "step": 2346
    },
    {
      "epoch": 1.78,
      "learning_rate": 3.891357791225416e-05,
      "loss": 0.8604,
      "step": 2347
    },
    {
      "epoch": 1.78,
      "learning_rate": 3.89088502269289e-05,
      "loss": 0.9203,
      "step": 2348
    },
    {
      "epoch": 1.78,
      "learning_rate": 3.890412254160363e-05,
      "loss": 0.9531,
      "step": 2349
    },
    {
      "epoch": 1.78,
      "learning_rate": 3.889939485627837e-05,
      "loss": 0.9767,
      "step": 2350
    },
    {
      "epoch": 1.78,
      "learning_rate": 3.8894667170953104e-05,
      "loss": 0.7706,
      "step": 2351
    },
    {
      "epoch": 1.78,
      "learning_rate": 3.888993948562784e-05,
      "loss": 0.8947,
      "step": 2352
    },
    {
      "epoch": 1.78,
      "learning_rate": 3.8885211800302575e-05,
      "loss": 0.9588,
      "step": 2353
    },
    {
      "epoch": 1.78,
      "learning_rate": 3.8880484114977314e-05,
      "loss": 0.8734,
      "step": 2354
    },
    {
      "epoch": 1.78,
      "learning_rate": 3.8875756429652046e-05,
      "loss": 0.8957,
      "step": 2355
    },
    {
      "epoch": 1.78,
      "learning_rate": 3.8871028744326785e-05,
      "loss": 0.863,
      "step": 2356
    },
    {
      "epoch": 1.78,
      "learning_rate": 3.886630105900152e-05,
      "loss": 0.8891,
      "step": 2357
    },
    {
      "epoch": 1.78,
      "learning_rate": 3.886157337367625e-05,
      "loss": 0.8193,
      "step": 2358
    },
    {
      "epoch": 1.78,
      "learning_rate": 3.885684568835099e-05,
      "loss": 0.9471,
      "step": 2359
    },
    {
      "epoch": 1.78,
      "learning_rate": 3.885211800302572e-05,
      "loss": 1.0082,
      "step": 2360
    },
    {
      "epoch": 1.79,
      "learning_rate": 3.884739031770045e-05,
      "loss": 0.9694,
      "step": 2361
    },
    {
      "epoch": 1.79,
      "learning_rate": 3.884266263237519e-05,
      "loss": 0.9028,
      "step": 2362
    },
    {
      "epoch": 1.79,
      "learning_rate": 3.883793494704992e-05,
      "loss": 0.8685,
      "step": 2363
    },
    {
      "epoch": 1.79,
      "learning_rate": 3.883320726172466e-05,
      "loss": 0.8862,
      "step": 2364
    },
    {
      "epoch": 1.79,
      "learning_rate": 3.8828479576399394e-05,
      "loss": 0.8535,
      "step": 2365
    },
    {
      "epoch": 1.79,
      "learning_rate": 3.882375189107413e-05,
      "loss": 0.898,
      "step": 2366
    },
    {
      "epoch": 1.79,
      "learning_rate": 3.8819024205748864e-05,
      "loss": 1.0786,
      "step": 2367
    },
    {
      "epoch": 1.79,
      "learning_rate": 3.88142965204236e-05,
      "loss": 0.9449,
      "step": 2368
    },
    {
      "epoch": 1.79,
      "learning_rate": 3.8809568835098335e-05,
      "loss": 0.8705,
      "step": 2369
    },
    {
      "epoch": 1.79,
      "learning_rate": 3.8804841149773074e-05,
      "loss": 0.9165,
      "step": 2370
    },
    {
      "epoch": 1.79,
      "learning_rate": 3.8800113464447806e-05,
      "loss": 0.8647,
      "step": 2371
    },
    {
      "epoch": 1.79,
      "learning_rate": 3.8795385779122545e-05,
      "loss": 0.8112,
      "step": 2372
    },
    {
      "epoch": 1.79,
      "learning_rate": 3.879065809379728e-05,
      "loss": 0.9507,
      "step": 2373
    },
    {
      "epoch": 1.8,
      "learning_rate": 3.8785930408472016e-05,
      "loss": 0.9207,
      "step": 2374
    },
    {
      "epoch": 1.8,
      "learning_rate": 3.878120272314675e-05,
      "loss": 0.8497,
      "step": 2375
    },
    {
      "epoch": 1.8,
      "learning_rate": 3.877647503782149e-05,
      "loss": 0.8733,
      "step": 2376
    },
    {
      "epoch": 1.8,
      "learning_rate": 3.877174735249622e-05,
      "loss": 0.9038,
      "step": 2377
    },
    {
      "epoch": 1.8,
      "learning_rate": 3.876701966717096e-05,
      "loss": 0.8416,
      "step": 2378
    },
    {
      "epoch": 1.8,
      "learning_rate": 3.876229198184569e-05,
      "loss": 0.898,
      "step": 2379
    },
    {
      "epoch": 1.8,
      "learning_rate": 3.875756429652043e-05,
      "loss": 1.0418,
      "step": 2380
    },
    {
      "epoch": 1.8,
      "learning_rate": 3.875283661119516e-05,
      "loss": 0.873,
      "step": 2381
    },
    {
      "epoch": 1.8,
      "learning_rate": 3.87481089258699e-05,
      "loss": 0.912,
      "step": 2382
    },
    {
      "epoch": 1.8,
      "learning_rate": 3.874338124054463e-05,
      "loss": 0.8446,
      "step": 2383
    },
    {
      "epoch": 1.8,
      "learning_rate": 3.873865355521937e-05,
      "loss": 0.8407,
      "step": 2384
    },
    {
      "epoch": 1.8,
      "learning_rate": 3.87339258698941e-05,
      "loss": 0.7955,
      "step": 2385
    },
    {
      "epoch": 1.8,
      "learning_rate": 3.8729198184568835e-05,
      "loss": 0.8848,
      "step": 2386
    },
    {
      "epoch": 1.81,
      "learning_rate": 3.8724470499243573e-05,
      "loss": 0.9486,
      "step": 2387
    },
    {
      "epoch": 1.81,
      "learning_rate": 3.8719742813918306e-05,
      "loss": 0.8909,
      "step": 2388
    },
    {
      "epoch": 1.81,
      "learning_rate": 3.871501512859304e-05,
      "loss": 0.8682,
      "step": 2389
    },
    {
      "epoch": 1.81,
      "learning_rate": 3.8710287443267776e-05,
      "loss": 0.8373,
      "step": 2390
    },
    {
      "epoch": 1.81,
      "learning_rate": 3.870555975794251e-05,
      "loss": 0.8412,
      "step": 2391
    },
    {
      "epoch": 1.81,
      "learning_rate": 3.870083207261725e-05,
      "loss": 0.8946,
      "step": 2392
    },
    {
      "epoch": 1.81,
      "learning_rate": 3.869610438729198e-05,
      "loss": 0.895,
      "step": 2393
    },
    {
      "epoch": 1.81,
      "learning_rate": 3.869137670196672e-05,
      "loss": 0.9033,
      "step": 2394
    },
    {
      "epoch": 1.81,
      "learning_rate": 3.868664901664145e-05,
      "loss": 0.8957,
      "step": 2395
    },
    {
      "epoch": 1.81,
      "learning_rate": 3.868192133131619e-05,
      "loss": 0.8317,
      "step": 2396
    },
    {
      "epoch": 1.81,
      "learning_rate": 3.867719364599092e-05,
      "loss": 0.9714,
      "step": 2397
    },
    {
      "epoch": 1.81,
      "learning_rate": 3.867246596066566e-05,
      "loss": 0.9066,
      "step": 2398
    },
    {
      "epoch": 1.81,
      "learning_rate": 3.866773827534039e-05,
      "loss": 0.8508,
      "step": 2399
    },
    {
      "epoch": 1.82,
      "learning_rate": 3.866301059001513e-05,
      "loss": 0.8759,
      "step": 2400
    },
    {
      "epoch": 1.82,
      "learning_rate": 3.865828290468986e-05,
      "loss": 0.8482,
      "step": 2401
    },
    {
      "epoch": 1.82,
      "learning_rate": 3.86535552193646e-05,
      "loss": 0.848,
      "step": 2402
    },
    {
      "epoch": 1.82,
      "learning_rate": 3.8648827534039334e-05,
      "loss": 0.8707,
      "step": 2403
    },
    {
      "epoch": 1.82,
      "learning_rate": 3.864409984871407e-05,
      "loss": 0.9242,
      "step": 2404
    },
    {
      "epoch": 1.82,
      "learning_rate": 3.8639372163388805e-05,
      "loss": 0.8377,
      "step": 2405
    },
    {
      "epoch": 1.82,
      "learning_rate": 3.8634644478063544e-05,
      "loss": 1.0221,
      "step": 2406
    },
    {
      "epoch": 1.82,
      "learning_rate": 3.862991679273828e-05,
      "loss": 0.9849,
      "step": 2407
    },
    {
      "epoch": 1.82,
      "learning_rate": 3.8625189107413014e-05,
      "loss": 0.7771,
      "step": 2408
    },
    {
      "epoch": 1.82,
      "learning_rate": 3.862046142208775e-05,
      "loss": 0.812,
      "step": 2409
    },
    {
      "epoch": 1.82,
      "learning_rate": 3.8615733736762485e-05,
      "loss": 0.951,
      "step": 2410
    },
    {
      "epoch": 1.82,
      "learning_rate": 3.861100605143722e-05,
      "loss": 0.8975,
      "step": 2411
    },
    {
      "epoch": 1.82,
      "learning_rate": 3.8606278366111956e-05,
      "loss": 0.8283,
      "step": 2412
    },
    {
      "epoch": 1.82,
      "learning_rate": 3.860155068078669e-05,
      "loss": 0.8697,
      "step": 2413
    },
    {
      "epoch": 1.83,
      "learning_rate": 3.859682299546142e-05,
      "loss": 0.8296,
      "step": 2414
    },
    {
      "epoch": 1.83,
      "learning_rate": 3.859209531013616e-05,
      "loss": 0.8851,
      "step": 2415
    },
    {
      "epoch": 1.83,
      "learning_rate": 3.858736762481089e-05,
      "loss": 0.8644,
      "step": 2416
    },
    {
      "epoch": 1.83,
      "learning_rate": 3.858263993948563e-05,
      "loss": 1.0253,
      "step": 2417
    },
    {
      "epoch": 1.83,
      "learning_rate": 3.857791225416036e-05,
      "loss": 1.0253,
      "step": 2418
    },
    {
      "epoch": 1.83,
      "learning_rate": 3.85731845688351e-05,
      "loss": 0.8448,
      "step": 2419
    },
    {
      "epoch": 1.83,
      "learning_rate": 3.856845688350983e-05,
      "loss": 0.9019,
      "step": 2420
    },
    {
      "epoch": 1.83,
      "learning_rate": 3.856372919818457e-05,
      "loss": 1.022,
      "step": 2421
    },
    {
      "epoch": 1.83,
      "learning_rate": 3.8559001512859304e-05,
      "loss": 0.8698,
      "step": 2422
    },
    {
      "epoch": 1.83,
      "learning_rate": 3.855427382753404e-05,
      "loss": 0.9729,
      "step": 2423
    },
    {
      "epoch": 1.83,
      "learning_rate": 3.8549546142208775e-05,
      "loss": 0.9221,
      "step": 2424
    },
    {
      "epoch": 1.83,
      "learning_rate": 3.8544818456883514e-05,
      "loss": 0.8646,
      "step": 2425
    },
    {
      "epoch": 1.83,
      "learning_rate": 3.8540090771558246e-05,
      "loss": 0.8858,
      "step": 2426
    },
    {
      "epoch": 1.84,
      "learning_rate": 3.8535363086232985e-05,
      "loss": 1.0001,
      "step": 2427
    },
    {
      "epoch": 1.84,
      "learning_rate": 3.853063540090772e-05,
      "loss": 0.8518,
      "step": 2428
    },
    {
      "epoch": 1.84,
      "learning_rate": 3.8525907715582455e-05,
      "loss": 0.8688,
      "step": 2429
    },
    {
      "epoch": 1.84,
      "learning_rate": 3.852118003025719e-05,
      "loss": 0.8571,
      "step": 2430
    },
    {
      "epoch": 1.84,
      "learning_rate": 3.8516452344931926e-05,
      "loss": 0.982,
      "step": 2431
    },
    {
      "epoch": 1.84,
      "learning_rate": 3.851172465960666e-05,
      "loss": 0.8598,
      "step": 2432
    },
    {
      "epoch": 1.84,
      "learning_rate": 3.85069969742814e-05,
      "loss": 0.8122,
      "step": 2433
    },
    {
      "epoch": 1.84,
      "learning_rate": 3.850226928895613e-05,
      "loss": 1.0525,
      "step": 2434
    },
    {
      "epoch": 1.84,
      "learning_rate": 3.849754160363087e-05,
      "loss": 0.8173,
      "step": 2435
    },
    {
      "epoch": 1.84,
      "learning_rate": 3.84928139183056e-05,
      "loss": 0.9339,
      "step": 2436
    },
    {
      "epoch": 1.84,
      "learning_rate": 3.848808623298034e-05,
      "loss": 0.8039,
      "step": 2437
    },
    {
      "epoch": 1.84,
      "learning_rate": 3.848335854765507e-05,
      "loss": 0.9341,
      "step": 2438
    },
    {
      "epoch": 1.84,
      "learning_rate": 3.84786308623298e-05,
      "loss": 0.9125,
      "step": 2439
    },
    {
      "epoch": 1.85,
      "learning_rate": 3.847390317700454e-05,
      "loss": 0.8651,
      "step": 2440
    },
    {
      "epoch": 1.85,
      "learning_rate": 3.8469175491679274e-05,
      "loss": 0.9425,
      "step": 2441
    },
    {
      "epoch": 1.85,
      "learning_rate": 3.8464447806354006e-05,
      "loss": 1.0273,
      "step": 2442
    },
    {
      "epoch": 1.85,
      "learning_rate": 3.8459720121028745e-05,
      "loss": 0.9794,
      "step": 2443
    },
    {
      "epoch": 1.85,
      "learning_rate": 3.845499243570348e-05,
      "loss": 0.809,
      "step": 2444
    },
    {
      "epoch": 1.85,
      "learning_rate": 3.8450264750378216e-05,
      "loss": 0.9284,
      "step": 2445
    },
    {
      "epoch": 1.85,
      "learning_rate": 3.844553706505295e-05,
      "loss": 0.7846,
      "step": 2446
    },
    {
      "epoch": 1.85,
      "learning_rate": 3.844080937972769e-05,
      "loss": 0.992,
      "step": 2447
    },
    {
      "epoch": 1.85,
      "learning_rate": 3.843608169440242e-05,
      "loss": 0.9588,
      "step": 2448
    },
    {
      "epoch": 1.85,
      "learning_rate": 3.843135400907716e-05,
      "loss": 1.0544,
      "step": 2449
    },
    {
      "epoch": 1.85,
      "learning_rate": 3.842662632375189e-05,
      "loss": 0.8294,
      "step": 2450
    },
    {
      "epoch": 1.85,
      "learning_rate": 3.842189863842663e-05,
      "loss": 0.8198,
      "step": 2451
    },
    {
      "epoch": 1.85,
      "learning_rate": 3.841717095310136e-05,
      "loss": 0.8269,
      "step": 2452
    },
    {
      "epoch": 1.86,
      "learning_rate": 3.84124432677761e-05,
      "loss": 0.8087,
      "step": 2453
    },
    {
      "epoch": 1.86,
      "learning_rate": 3.840771558245083e-05,
      "loss": 0.8568,
      "step": 2454
    },
    {
      "epoch": 1.86,
      "learning_rate": 3.840298789712557e-05,
      "loss": 0.8684,
      "step": 2455
    },
    {
      "epoch": 1.86,
      "learning_rate": 3.83982602118003e-05,
      "loss": 0.8012,
      "step": 2456
    },
    {
      "epoch": 1.86,
      "learning_rate": 3.839353252647504e-05,
      "loss": 0.9238,
      "step": 2457
    },
    {
      "epoch": 1.86,
      "learning_rate": 3.838880484114977e-05,
      "loss": 0.8651,
      "step": 2458
    },
    {
      "epoch": 1.86,
      "learning_rate": 3.838407715582451e-05,
      "loss": 0.8608,
      "step": 2459
    },
    {
      "epoch": 1.86,
      "learning_rate": 3.8379349470499244e-05,
      "loss": 0.9271,
      "step": 2460
    },
    {
      "epoch": 1.86,
      "learning_rate": 3.837462178517398e-05,
      "loss": 0.7744,
      "step": 2461
    },
    {
      "epoch": 1.86,
      "learning_rate": 3.8369894099848715e-05,
      "loss": 0.91,
      "step": 2462
    },
    {
      "epoch": 1.86,
      "learning_rate": 3.8365166414523454e-05,
      "loss": 0.9263,
      "step": 2463
    },
    {
      "epoch": 1.86,
      "learning_rate": 3.8360438729198186e-05,
      "loss": 0.9002,
      "step": 2464
    },
    {
      "epoch": 1.86,
      "learning_rate": 3.8355711043872925e-05,
      "loss": 0.9796,
      "step": 2465
    },
    {
      "epoch": 1.87,
      "learning_rate": 3.835098335854766e-05,
      "loss": 0.8395,
      "step": 2466
    },
    {
      "epoch": 1.87,
      "learning_rate": 3.834625567322239e-05,
      "loss": 0.8268,
      "step": 2467
    },
    {
      "epoch": 1.87,
      "learning_rate": 3.834152798789713e-05,
      "loss": 0.8788,
      "step": 2468
    },
    {
      "epoch": 1.87,
      "learning_rate": 3.833680030257186e-05,
      "loss": 0.7903,
      "step": 2469
    },
    {
      "epoch": 1.87,
      "learning_rate": 3.83320726172466e-05,
      "loss": 0.8489,
      "step": 2470
    },
    {
      "epoch": 1.87,
      "learning_rate": 3.832734493192133e-05,
      "loss": 1.023,
      "step": 2471
    },
    {
      "epoch": 1.87,
      "learning_rate": 3.832261724659607e-05,
      "loss": 0.8853,
      "step": 2472
    },
    {
      "epoch": 1.87,
      "learning_rate": 3.83178895612708e-05,
      "loss": 0.8803,
      "step": 2473
    },
    {
      "epoch": 1.87,
      "learning_rate": 3.831316187594554e-05,
      "loss": 0.9594,
      "step": 2474
    },
    {
      "epoch": 1.87,
      "learning_rate": 3.830843419062027e-05,
      "loss": 0.8684,
      "step": 2475
    },
    {
      "epoch": 1.87,
      "learning_rate": 3.830370650529501e-05,
      "loss": 0.8732,
      "step": 2476
    },
    {
      "epoch": 1.87,
      "learning_rate": 3.829897881996974e-05,
      "loss": 0.975,
      "step": 2477
    },
    {
      "epoch": 1.87,
      "learning_rate": 3.829425113464448e-05,
      "loss": 0.8149,
      "step": 2478
    },
    {
      "epoch": 1.87,
      "learning_rate": 3.8289523449319214e-05,
      "loss": 0.9037,
      "step": 2479
    },
    {
      "epoch": 1.88,
      "learning_rate": 3.828479576399395e-05,
      "loss": 0.8342,
      "step": 2480
    },
    {
      "epoch": 1.88,
      "learning_rate": 3.8280068078668685e-05,
      "loss": 0.9767,
      "step": 2481
    },
    {
      "epoch": 1.88,
      "learning_rate": 3.8275340393343424e-05,
      "loss": 0.8852,
      "step": 2482
    },
    {
      "epoch": 1.88,
      "learning_rate": 3.8270612708018156e-05,
      "loss": 0.8204,
      "step": 2483
    },
    {
      "epoch": 1.88,
      "learning_rate": 3.8265885022692895e-05,
      "loss": 0.9626,
      "step": 2484
    },
    {
      "epoch": 1.88,
      "learning_rate": 3.826115733736763e-05,
      "loss": 0.9323,
      "step": 2485
    },
    {
      "epoch": 1.88,
      "learning_rate": 3.8256429652042366e-05,
      "loss": 0.9184,
      "step": 2486
    },
    {
      "epoch": 1.88,
      "learning_rate": 3.82517019667171e-05,
      "loss": 0.8563,
      "step": 2487
    },
    {
      "epoch": 1.88,
      "learning_rate": 3.824697428139184e-05,
      "loss": 0.7796,
      "step": 2488
    },
    {
      "epoch": 1.88,
      "learning_rate": 3.824224659606657e-05,
      "loss": 0.8487,
      "step": 2489
    },
    {
      "epoch": 1.88,
      "learning_rate": 3.823751891074131e-05,
      "loss": 0.9392,
      "step": 2490
    },
    {
      "epoch": 1.88,
      "learning_rate": 3.823279122541604e-05,
      "loss": 0.8732,
      "step": 2491
    },
    {
      "epoch": 1.88,
      "learning_rate": 3.822806354009077e-05,
      "loss": 0.8334,
      "step": 2492
    },
    {
      "epoch": 1.89,
      "learning_rate": 3.822333585476551e-05,
      "loss": 0.9112,
      "step": 2493
    },
    {
      "epoch": 1.89,
      "learning_rate": 3.821860816944024e-05,
      "loss": 0.8823,
      "step": 2494
    },
    {
      "epoch": 1.89,
      "learning_rate": 3.8213880484114975e-05,
      "loss": 0.7964,
      "step": 2495
    },
    {
      "epoch": 1.89,
      "learning_rate": 3.8209152798789713e-05,
      "loss": 0.8245,
      "step": 2496
    },
    {
      "epoch": 1.89,
      "learning_rate": 3.8204425113464446e-05,
      "loss": 0.8324,
      "step": 2497
    },
    {
      "epoch": 1.89,
      "learning_rate": 3.8199697428139184e-05,
      "loss": 0.8999,
      "step": 2498
    },
    {
      "epoch": 1.89,
      "learning_rate": 3.8194969742813916e-05,
      "loss": 1.0321,
      "step": 2499
    },
    {
      "epoch": 1.89,
      "learning_rate": 3.8190242057488655e-05,
      "loss": 0.9642,
      "step": 2500
    },
    {
      "epoch": 1.89,
      "learning_rate": 3.818551437216339e-05,
      "loss": 1.0115,
      "step": 2501
    },
    {
      "epoch": 1.89,
      "learning_rate": 3.8180786686838126e-05,
      "loss": 0.9873,
      "step": 2502
    },
    {
      "epoch": 1.89,
      "learning_rate": 3.817605900151286e-05,
      "loss": 0.9753,
      "step": 2503
    },
    {
      "epoch": 1.89,
      "learning_rate": 3.81713313161876e-05,
      "loss": 0.9715,
      "step": 2504
    },
    {
      "epoch": 1.89,
      "learning_rate": 3.816660363086233e-05,
      "loss": 0.9719,
      "step": 2505
    },
    {
      "epoch": 1.9,
      "learning_rate": 3.816187594553707e-05,
      "loss": 0.8762,
      "step": 2506
    },
    {
      "epoch": 1.9,
      "learning_rate": 3.81571482602118e-05,
      "loss": 0.9852,
      "step": 2507
    },
    {
      "epoch": 1.9,
      "learning_rate": 3.815242057488654e-05,
      "loss": 0.8357,
      "step": 2508
    },
    {
      "epoch": 1.9,
      "learning_rate": 3.814769288956127e-05,
      "loss": 0.9927,
      "step": 2509
    },
    {
      "epoch": 1.9,
      "learning_rate": 3.814296520423601e-05,
      "loss": 0.887,
      "step": 2510
    },
    {
      "epoch": 1.9,
      "learning_rate": 3.813823751891074e-05,
      "loss": 0.9168,
      "step": 2511
    },
    {
      "epoch": 1.9,
      "learning_rate": 3.813350983358548e-05,
      "loss": 0.9584,
      "step": 2512
    },
    {
      "epoch": 1.9,
      "learning_rate": 3.812878214826021e-05,
      "loss": 0.9114,
      "step": 2513
    },
    {
      "epoch": 1.9,
      "learning_rate": 3.812405446293495e-05,
      "loss": 0.8167,
      "step": 2514
    },
    {
      "epoch": 1.9,
      "learning_rate": 3.8119326777609684e-05,
      "loss": 0.8853,
      "step": 2515
    },
    {
      "epoch": 1.9,
      "learning_rate": 3.811459909228442e-05,
      "loss": 0.9575,
      "step": 2516
    },
    {
      "epoch": 1.9,
      "learning_rate": 3.8109871406959154e-05,
      "loss": 0.9043,
      "step": 2517
    },
    {
      "epoch": 1.9,
      "learning_rate": 3.810514372163389e-05,
      "loss": 0.8134,
      "step": 2518
    },
    {
      "epoch": 1.91,
      "learning_rate": 3.8100416036308625e-05,
      "loss": 0.9244,
      "step": 2519
    },
    {
      "epoch": 1.91,
      "learning_rate": 3.809568835098336e-05,
      "loss": 0.9625,
      "step": 2520
    },
    {
      "epoch": 1.91,
      "learning_rate": 3.8090960665658096e-05,
      "loss": 0.7795,
      "step": 2521
    },
    {
      "epoch": 1.91,
      "learning_rate": 3.808623298033283e-05,
      "loss": 0.8382,
      "step": 2522
    },
    {
      "epoch": 1.91,
      "learning_rate": 3.808150529500756e-05,
      "loss": 0.8309,
      "step": 2523
    },
    {
      "epoch": 1.91,
      "learning_rate": 3.80767776096823e-05,
      "loss": 0.8953,
      "step": 2524
    },
    {
      "epoch": 1.91,
      "learning_rate": 3.807204992435703e-05,
      "loss": 0.6905,
      "step": 2525
    },
    {
      "epoch": 1.91,
      "learning_rate": 3.806732223903177e-05,
      "loss": 0.8702,
      "step": 2526
    },
    {
      "epoch": 1.91,
      "learning_rate": 3.806259455370651e-05,
      "loss": 0.914,
      "step": 2527
    },
    {
      "epoch": 1.91,
      "learning_rate": 3.805786686838124e-05,
      "loss": 0.8983,
      "step": 2528
    },
    {
      "epoch": 1.91,
      "learning_rate": 3.805313918305598e-05,
      "loss": 0.9944,
      "step": 2529
    },
    {
      "epoch": 1.91,
      "learning_rate": 3.804841149773071e-05,
      "loss": 0.8786,
      "step": 2530
    },
    {
      "epoch": 1.91,
      "learning_rate": 3.804368381240545e-05,
      "loss": 0.879,
      "step": 2531
    },
    {
      "epoch": 1.91,
      "learning_rate": 3.803895612708018e-05,
      "loss": 0.9395,
      "step": 2532
    },
    {
      "epoch": 1.92,
      "learning_rate": 3.803422844175492e-05,
      "loss": 0.9653,
      "step": 2533
    },
    {
      "epoch": 1.92,
      "learning_rate": 3.8029500756429654e-05,
      "loss": 0.8039,
      "step": 2534
    },
    {
      "epoch": 1.92,
      "learning_rate": 3.802477307110439e-05,
      "loss": 0.9099,
      "step": 2535
    },
    {
      "epoch": 1.92,
      "learning_rate": 3.8020045385779125e-05,
      "loss": 1.0041,
      "step": 2536
    },
    {
      "epoch": 1.92,
      "learning_rate": 3.8015317700453863e-05,
      "loss": 0.9962,
      "step": 2537
    },
    {
      "epoch": 1.92,
      "learning_rate": 3.8010590015128596e-05,
      "loss": 0.8482,
      "step": 2538
    },
    {
      "epoch": 1.92,
      "learning_rate": 3.8005862329803334e-05,
      "loss": 0.8659,
      "step": 2539
    },
    {
      "epoch": 1.92,
      "learning_rate": 3.8001134644478066e-05,
      "loss": 1.0008,
      "step": 2540
    },
    {
      "epoch": 1.92,
      "learning_rate": 3.7996406959152805e-05,
      "loss": 0.7769,
      "step": 2541
    },
    {
      "epoch": 1.92,
      "learning_rate": 3.799167927382754e-05,
      "loss": 0.9848,
      "step": 2542
    },
    {
      "epoch": 1.92,
      "learning_rate": 3.7986951588502276e-05,
      "loss": 0.884,
      "step": 2543
    },
    {
      "epoch": 1.92,
      "learning_rate": 3.798222390317701e-05,
      "loss": 0.8367,
      "step": 2544
    },
    {
      "epoch": 1.92,
      "learning_rate": 3.797749621785174e-05,
      "loss": 0.8446,
      "step": 2545
    },
    {
      "epoch": 1.93,
      "learning_rate": 3.797276853252648e-05,
      "loss": 0.8366,
      "step": 2546
    },
    {
      "epoch": 1.93,
      "learning_rate": 3.796804084720121e-05,
      "loss": 0.8998,
      "step": 2547
    },
    {
      "epoch": 1.93,
      "learning_rate": 3.796331316187594e-05,
      "loss": 0.8534,
      "step": 2548
    },
    {
      "epoch": 1.93,
      "learning_rate": 3.795858547655068e-05,
      "loss": 0.8518,
      "step": 2549
    },
    {
      "epoch": 1.93,
      "learning_rate": 3.7953857791225414e-05,
      "loss": 0.9408,
      "step": 2550
    },
    {
      "epoch": 1.93,
      "learning_rate": 3.794913010590015e-05,
      "loss": 0.8578,
      "step": 2551
    },
    {
      "epoch": 1.93,
      "learning_rate": 3.7944402420574885e-05,
      "loss": 0.9664,
      "step": 2552
    },
    {
      "epoch": 1.93,
      "learning_rate": 3.7939674735249624e-05,
      "loss": 0.9353,
      "step": 2553
    },
    {
      "epoch": 1.93,
      "learning_rate": 3.7934947049924356e-05,
      "loss": 0.7891,
      "step": 2554
    },
    {
      "epoch": 1.93,
      "learning_rate": 3.7930219364599095e-05,
      "loss": 1.0583,
      "step": 2555
    },
    {
      "epoch": 1.93,
      "learning_rate": 3.792549167927383e-05,
      "loss": 0.9855,
      "step": 2556
    },
    {
      "epoch": 1.93,
      "learning_rate": 3.7920763993948566e-05,
      "loss": 0.924,
      "step": 2557
    },
    {
      "epoch": 1.93,
      "learning_rate": 3.79160363086233e-05,
      "loss": 0.9281,
      "step": 2558
    },
    {
      "epoch": 1.94,
      "learning_rate": 3.7911308623298037e-05,
      "loss": 0.8857,
      "step": 2559
    },
    {
      "epoch": 1.94,
      "learning_rate": 3.790658093797277e-05,
      "loss": 0.8914,
      "step": 2560
    },
    {
      "epoch": 1.94,
      "learning_rate": 3.790185325264751e-05,
      "loss": 0.9533,
      "step": 2561
    },
    {
      "epoch": 1.94,
      "learning_rate": 3.789712556732224e-05,
      "loss": 0.8589,
      "step": 2562
    },
    {
      "epoch": 1.94,
      "learning_rate": 3.789239788199698e-05,
      "loss": 0.8606,
      "step": 2563
    },
    {
      "epoch": 1.94,
      "learning_rate": 3.788767019667171e-05,
      "loss": 0.8953,
      "step": 2564
    },
    {
      "epoch": 1.94,
      "learning_rate": 3.788294251134645e-05,
      "loss": 0.9327,
      "step": 2565
    },
    {
      "epoch": 1.94,
      "learning_rate": 3.787821482602118e-05,
      "loss": 0.9071,
      "step": 2566
    },
    {
      "epoch": 1.94,
      "learning_rate": 3.787348714069592e-05,
      "loss": 0.8489,
      "step": 2567
    },
    {
      "epoch": 1.94,
      "learning_rate": 3.786875945537065e-05,
      "loss": 0.9125,
      "step": 2568
    },
    {
      "epoch": 1.94,
      "learning_rate": 3.786403177004539e-05,
      "loss": 0.9674,
      "step": 2569
    },
    {
      "epoch": 1.94,
      "learning_rate": 3.785930408472012e-05,
      "loss": 0.8997,
      "step": 2570
    },
    {
      "epoch": 1.94,
      "learning_rate": 3.785457639939486e-05,
      "loss": 0.9253,
      "step": 2571
    },
    {
      "epoch": 1.95,
      "learning_rate": 3.7849848714069594e-05,
      "loss": 0.8791,
      "step": 2572
    },
    {
      "epoch": 1.95,
      "learning_rate": 3.7845121028744326e-05,
      "loss": 0.8969,
      "step": 2573
    },
    {
      "epoch": 1.95,
      "learning_rate": 3.7840393343419065e-05,
      "loss": 0.898,
      "step": 2574
    },
    {
      "epoch": 1.95,
      "learning_rate": 3.78356656580938e-05,
      "loss": 0.8551,
      "step": 2575
    },
    {
      "epoch": 1.95,
      "learning_rate": 3.783093797276853e-05,
      "loss": 0.8829,
      "step": 2576
    },
    {
      "epoch": 1.95,
      "learning_rate": 3.782621028744327e-05,
      "loss": 0.8835,
      "step": 2577
    },
    {
      "epoch": 1.95,
      "learning_rate": 3.7821482602118e-05,
      "loss": 0.8119,
      "step": 2578
    },
    {
      "epoch": 1.95,
      "learning_rate": 3.781675491679274e-05,
      "loss": 0.9587,
      "step": 2579
    },
    {
      "epoch": 1.95,
      "learning_rate": 3.781202723146747e-05,
      "loss": 0.969,
      "step": 2580
    },
    {
      "epoch": 1.95,
      "learning_rate": 3.780729954614221e-05,
      "loss": 0.8306,
      "step": 2581
    },
    {
      "epoch": 1.95,
      "learning_rate": 3.780257186081694e-05,
      "loss": 0.9675,
      "step": 2582
    },
    {
      "epoch": 1.95,
      "learning_rate": 3.779784417549168e-05,
      "loss": 0.9041,
      "step": 2583
    },
    {
      "epoch": 1.95,
      "learning_rate": 3.779311649016641e-05,
      "loss": 0.8593,
      "step": 2584
    },
    {
      "epoch": 1.96,
      "learning_rate": 3.778838880484115e-05,
      "loss": 0.9429,
      "step": 2585
    },
    {
      "epoch": 1.96,
      "learning_rate": 3.778366111951589e-05,
      "loss": 0.9883,
      "step": 2586
    },
    {
      "epoch": 1.96,
      "learning_rate": 3.777893343419062e-05,
      "loss": 0.8948,
      "step": 2587
    },
    {
      "epoch": 1.96,
      "learning_rate": 3.777420574886536e-05,
      "loss": 0.8622,
      "step": 2588
    },
    {
      "epoch": 1.96,
      "learning_rate": 3.776947806354009e-05,
      "loss": 0.8066,
      "step": 2589
    },
    {
      "epoch": 1.96,
      "learning_rate": 3.776475037821483e-05,
      "loss": 0.8887,
      "step": 2590
    },
    {
      "epoch": 1.96,
      "learning_rate": 3.7760022692889564e-05,
      "loss": 0.8134,
      "step": 2591
    },
    {
      "epoch": 1.96,
      "learning_rate": 3.77552950075643e-05,
      "loss": 0.8332,
      "step": 2592
    },
    {
      "epoch": 1.96,
      "learning_rate": 3.7750567322239035e-05,
      "loss": 0.8395,
      "step": 2593
    },
    {
      "epoch": 1.96,
      "learning_rate": 3.7745839636913774e-05,
      "loss": 0.873,
      "step": 2594
    },
    {
      "epoch": 1.96,
      "learning_rate": 3.7741111951588506e-05,
      "loss": 0.9161,
      "step": 2595
    },
    {
      "epoch": 1.96,
      "learning_rate": 3.7736384266263245e-05,
      "loss": 0.8624,
      "step": 2596
    },
    {
      "epoch": 1.96,
      "learning_rate": 3.773165658093798e-05,
      "loss": 0.8045,
      "step": 2597
    },
    {
      "epoch": 1.96,
      "learning_rate": 3.772692889561271e-05,
      "loss": 0.8501,
      "step": 2598
    },
    {
      "epoch": 1.97,
      "learning_rate": 3.772220121028745e-05,
      "loss": 0.9476,
      "step": 2599
    },
    {
      "epoch": 1.97,
      "learning_rate": 3.771747352496218e-05,
      "loss": 0.902,
      "step": 2600
    },
    {
      "epoch": 1.97,
      "learning_rate": 3.771274583963691e-05,
      "loss": 0.8581,
      "step": 2601
    },
    {
      "epoch": 1.97,
      "learning_rate": 3.770801815431165e-05,
      "loss": 0.9091,
      "step": 2602
    },
    {
      "epoch": 1.97,
      "learning_rate": 3.770329046898638e-05,
      "loss": 0.9239,
      "step": 2603
    },
    {
      "epoch": 1.97,
      "learning_rate": 3.769856278366112e-05,
      "loss": 0.9306,
      "step": 2604
    },
    {
      "epoch": 1.97,
      "learning_rate": 3.7693835098335854e-05,
      "loss": 0.9677,
      "step": 2605
    },
    {
      "epoch": 1.97,
      "learning_rate": 3.768910741301059e-05,
      "loss": 0.888,
      "step": 2606
    },
    {
      "epoch": 1.97,
      "learning_rate": 3.7684379727685324e-05,
      "loss": 0.8035,
      "step": 2607
    },
    {
      "epoch": 1.97,
      "learning_rate": 3.767965204236006e-05,
      "loss": 0.8222,
      "step": 2608
    },
    {
      "epoch": 1.97,
      "learning_rate": 3.7674924357034795e-05,
      "loss": 0.8738,
      "step": 2609
    },
    {
      "epoch": 1.97,
      "learning_rate": 3.7670196671709534e-05,
      "loss": 0.7585,
      "step": 2610
    },
    {
      "epoch": 1.97,
      "learning_rate": 3.7665468986384266e-05,
      "loss": 0.8723,
      "step": 2611
    },
    {
      "epoch": 1.98,
      "learning_rate": 3.7660741301059005e-05,
      "loss": 0.9532,
      "step": 2612
    },
    {
      "epoch": 1.98,
      "learning_rate": 3.765601361573374e-05,
      "loss": 0.9143,
      "step": 2613
    },
    {
      "epoch": 1.98,
      "learning_rate": 3.7651285930408476e-05,
      "loss": 0.94,
      "step": 2614
    },
    {
      "epoch": 1.98,
      "learning_rate": 3.764655824508321e-05,
      "loss": 0.8629,
      "step": 2615
    },
    {
      "epoch": 1.98,
      "learning_rate": 3.764183055975795e-05,
      "loss": 0.9791,
      "step": 2616
    },
    {
      "epoch": 1.98,
      "learning_rate": 3.763710287443268e-05,
      "loss": 0.8596,
      "step": 2617
    },
    {
      "epoch": 1.98,
      "learning_rate": 3.763237518910742e-05,
      "loss": 0.853,
      "step": 2618
    },
    {
      "epoch": 1.98,
      "learning_rate": 3.762764750378215e-05,
      "loss": 0.8764,
      "step": 2619
    },
    {
      "epoch": 1.98,
      "learning_rate": 3.762291981845689e-05,
      "loss": 0.9504,
      "step": 2620
    },
    {
      "epoch": 1.98,
      "learning_rate": 3.761819213313162e-05,
      "loss": 0.9352,
      "step": 2621
    },
    {
      "epoch": 1.98,
      "learning_rate": 3.761346444780636e-05,
      "loss": 0.9188,
      "step": 2622
    },
    {
      "epoch": 1.98,
      "learning_rate": 3.760873676248109e-05,
      "loss": 0.9154,
      "step": 2623
    },
    {
      "epoch": 1.98,
      "learning_rate": 3.760400907715583e-05,
      "loss": 0.8201,
      "step": 2624
    },
    {
      "epoch": 1.99,
      "learning_rate": 3.759928139183056e-05,
      "loss": 0.9663,
      "step": 2625
    },
    {
      "epoch": 1.99,
      "learning_rate": 3.7594553706505295e-05,
      "loss": 0.8462,
      "step": 2626
    },
    {
      "epoch": 1.99,
      "learning_rate": 3.758982602118003e-05,
      "loss": 0.8624,
      "step": 2627
    },
    {
      "epoch": 1.99,
      "learning_rate": 3.7585098335854765e-05,
      "loss": 1.0373,
      "step": 2628
    },
    {
      "epoch": 1.99,
      "learning_rate": 3.75803706505295e-05,
      "loss": 0.887,
      "step": 2629
    },
    {
      "epoch": 1.99,
      "learning_rate": 3.7575642965204236e-05,
      "loss": 0.9287,
      "step": 2630
    },
    {
      "epoch": 1.99,
      "learning_rate": 3.757091527987897e-05,
      "loss": 0.9733,
      "step": 2631
    },
    {
      "epoch": 1.99,
      "learning_rate": 3.756618759455371e-05,
      "loss": 0.8364,
      "step": 2632
    },
    {
      "epoch": 1.99,
      "learning_rate": 3.756145990922844e-05,
      "loss": 0.8541,
      "step": 2633
    },
    {
      "epoch": 1.99,
      "learning_rate": 3.755673222390318e-05,
      "loss": 0.8262,
      "step": 2634
    },
    {
      "epoch": 1.99,
      "learning_rate": 3.755200453857791e-05,
      "loss": 0.8964,
      "step": 2635
    },
    {
      "epoch": 1.99,
      "learning_rate": 3.754727685325265e-05,
      "loss": 0.7654,
      "step": 2636
    },
    {
      "epoch": 1.99,
      "learning_rate": 3.754254916792738e-05,
      "loss": 0.9505,
      "step": 2637
    },
    {
      "epoch": 2.0,
      "learning_rate": 3.753782148260212e-05,
      "loss": 1.006,
      "step": 2638
    },
    {
      "epoch": 2.0,
      "learning_rate": 3.753309379727685e-05,
      "loss": 0.8699,
      "step": 2639
    },
    {
      "epoch": 2.0,
      "learning_rate": 3.752836611195159e-05,
      "loss": 0.9399,
      "step": 2640
    },
    {
      "epoch": 2.0,
      "learning_rate": 3.752363842662632e-05,
      "loss": 0.8242,
      "step": 2641
    },
    {
      "epoch": 2.0,
      "learning_rate": 3.751891074130106e-05,
      "loss": 1.0122,
      "step": 2642
    },
    {
      "epoch": 2.0,
      "learning_rate": 3.7514183055975794e-05,
      "loss": 0.9879,
      "step": 2643
    },
    {
      "epoch": 2.0,
      "learning_rate": 3.750945537065053e-05,
      "loss": 0.8652,
      "step": 2644
    },
    {
      "epoch": 2.0,
      "learning_rate": 3.7504727685325265e-05,
      "loss": 0.6463,
      "step": 2645
    },
    {
      "epoch": 2.0,
      "learning_rate": 3.7500000000000003e-05,
      "loss": 0.2642,
      "step": 2646
    },
    {
      "epoch": 2.0,
      "learning_rate": 3.749527231467474e-05,
      "loss": 0.2791,
      "step": 2647
    },
    {
      "epoch": 2.0,
      "learning_rate": 3.7490544629349474e-05,
      "loss": 0.2656,
      "step": 2648
    },
    {
      "epoch": 2.0,
      "learning_rate": 3.748581694402421e-05,
      "loss": 0.2777,
      "step": 2649
    },
    {
      "epoch": 2.0,
      "learning_rate": 3.7481089258698945e-05,
      "loss": 0.2345,
      "step": 2650
    },
    {
      "epoch": 2.0,
      "learning_rate": 3.747636157337368e-05,
      "loss": 0.2687,
      "step": 2651
    },
    {
      "epoch": 2.01,
      "learning_rate": 3.7471633888048416e-05,
      "loss": 0.283,
      "step": 2652
    },
    {
      "epoch": 2.01,
      "learning_rate": 3.746690620272315e-05,
      "loss": 0.2413,
      "step": 2653
    },
    {
      "epoch": 2.01,
      "learning_rate": 3.746217851739788e-05,
      "loss": 0.2339,
      "step": 2654
    },
    {
      "epoch": 2.01,
      "learning_rate": 3.745745083207262e-05,
      "loss": 0.2605,
      "step": 2655
    },
    {
      "epoch": 2.01,
      "learning_rate": 3.745272314674735e-05,
      "loss": 0.2916,
      "step": 2656
    },
    {
      "epoch": 2.01,
      "learning_rate": 3.744799546142209e-05,
      "loss": 0.2188,
      "step": 2657
    },
    {
      "epoch": 2.01,
      "learning_rate": 3.744326777609682e-05,
      "loss": 0.2208,
      "step": 2658
    },
    {
      "epoch": 2.01,
      "learning_rate": 3.743854009077156e-05,
      "loss": 0.2583,
      "step": 2659
    },
    {
      "epoch": 2.01,
      "learning_rate": 3.743381240544629e-05,
      "loss": 0.3086,
      "step": 2660
    },
    {
      "epoch": 2.01,
      "learning_rate": 3.742908472012103e-05,
      "loss": 0.2537,
      "step": 2661
    },
    {
      "epoch": 2.01,
      "learning_rate": 3.7424357034795764e-05,
      "loss": 0.2863,
      "step": 2662
    },
    {
      "epoch": 2.01,
      "learning_rate": 3.74196293494705e-05,
      "loss": 0.2327,
      "step": 2663
    },
    {
      "epoch": 2.01,
      "learning_rate": 3.7414901664145235e-05,
      "loss": 0.2863,
      "step": 2664
    },
    {
      "epoch": 2.02,
      "learning_rate": 3.7410173978819974e-05,
      "loss": 0.2675,
      "step": 2665
    },
    {
      "epoch": 2.02,
      "learning_rate": 3.7405446293494706e-05,
      "loss": 0.2337,
      "step": 2666
    },
    {
      "epoch": 2.02,
      "learning_rate": 3.7400718608169444e-05,
      "loss": 0.2477,
      "step": 2667
    },
    {
      "epoch": 2.02,
      "learning_rate": 3.7395990922844177e-05,
      "loss": 0.2361,
      "step": 2668
    },
    {
      "epoch": 2.02,
      "learning_rate": 3.7391263237518915e-05,
      "loss": 0.2534,
      "step": 2669
    },
    {
      "epoch": 2.02,
      "learning_rate": 3.738653555219365e-05,
      "loss": 0.233,
      "step": 2670
    },
    {
      "epoch": 2.02,
      "learning_rate": 3.7381807866868386e-05,
      "loss": 0.1913,
      "step": 2671
    },
    {
      "epoch": 2.02,
      "learning_rate": 3.737708018154312e-05,
      "loss": 0.2104,
      "step": 2672
    },
    {
      "epoch": 2.02,
      "learning_rate": 3.737235249621786e-05,
      "loss": 0.2621,
      "step": 2673
    },
    {
      "epoch": 2.02,
      "learning_rate": 3.736762481089259e-05,
      "loss": 0.2314,
      "step": 2674
    },
    {
      "epoch": 2.02,
      "learning_rate": 3.736289712556733e-05,
      "loss": 0.2897,
      "step": 2675
    },
    {
      "epoch": 2.02,
      "learning_rate": 3.735816944024206e-05,
      "loss": 0.2124,
      "step": 2676
    },
    {
      "epoch": 2.02,
      "learning_rate": 3.73534417549168e-05,
      "loss": 0.2044,
      "step": 2677
    },
    {
      "epoch": 2.03,
      "learning_rate": 3.734871406959153e-05,
      "loss": 0.2211,
      "step": 2678
    },
    {
      "epoch": 2.03,
      "learning_rate": 3.734398638426626e-05,
      "loss": 0.2896,
      "step": 2679
    },
    {
      "epoch": 2.03,
      "learning_rate": 3.7339258698941e-05,
      "loss": 0.2022,
      "step": 2680
    },
    {
      "epoch": 2.03,
      "learning_rate": 3.7334531013615734e-05,
      "loss": 0.2613,
      "step": 2681
    },
    {
      "epoch": 2.03,
      "learning_rate": 3.7329803328290466e-05,
      "loss": 0.2896,
      "step": 2682
    },
    {
      "epoch": 2.03,
      "learning_rate": 3.7325075642965205e-05,
      "loss": 0.2243,
      "step": 2683
    },
    {
      "epoch": 2.03,
      "learning_rate": 3.732034795763994e-05,
      "loss": 0.2314,
      "step": 2684
    },
    {
      "epoch": 2.03,
      "learning_rate": 3.7315620272314676e-05,
      "loss": 0.2944,
      "step": 2685
    },
    {
      "epoch": 2.03,
      "learning_rate": 3.731089258698941e-05,
      "loss": 0.2411,
      "step": 2686
    },
    {
      "epoch": 2.03,
      "learning_rate": 3.730616490166415e-05,
      "loss": 0.2105,
      "step": 2687
    },
    {
      "epoch": 2.03,
      "learning_rate": 3.730143721633888e-05,
      "loss": 0.2415,
      "step": 2688
    },
    {
      "epoch": 2.03,
      "learning_rate": 3.729670953101362e-05,
      "loss": 0.2254,
      "step": 2689
    },
    {
      "epoch": 2.03,
      "learning_rate": 3.729198184568835e-05,
      "loss": 0.2075,
      "step": 2690
    },
    {
      "epoch": 2.04,
      "learning_rate": 3.728725416036309e-05,
      "loss": 0.2243,
      "step": 2691
    },
    {
      "epoch": 2.04,
      "learning_rate": 3.728252647503782e-05,
      "loss": 0.2349,
      "step": 2692
    },
    {
      "epoch": 2.04,
      "learning_rate": 3.727779878971256e-05,
      "loss": 0.2158,
      "step": 2693
    },
    {
      "epoch": 2.04,
      "learning_rate": 3.727307110438729e-05,
      "loss": 0.2504,
      "step": 2694
    },
    {
      "epoch": 2.04,
      "learning_rate": 3.726834341906203e-05,
      "loss": 0.2041,
      "step": 2695
    },
    {
      "epoch": 2.04,
      "learning_rate": 3.726361573373676e-05,
      "loss": 0.2722,
      "step": 2696
    },
    {
      "epoch": 2.04,
      "learning_rate": 3.72588880484115e-05,
      "loss": 0.2318,
      "step": 2697
    },
    {
      "epoch": 2.04,
      "learning_rate": 3.725416036308623e-05,
      "loss": 0.2924,
      "step": 2698
    },
    {
      "epoch": 2.04,
      "learning_rate": 3.724943267776097e-05,
      "loss": 0.2528,
      "step": 2699
    },
    {
      "epoch": 2.04,
      "learning_rate": 3.7244704992435704e-05,
      "loss": 0.2876,
      "step": 2700
    },
    {
      "epoch": 2.04,
      "learning_rate": 3.723997730711044e-05,
      "loss": 0.2407,
      "step": 2701
    },
    {
      "epoch": 2.04,
      "learning_rate": 3.7235249621785175e-05,
      "loss": 0.2678,
      "step": 2702
    },
    {
      "epoch": 2.04,
      "learning_rate": 3.7230521936459914e-05,
      "loss": 0.2925,
      "step": 2703
    },
    {
      "epoch": 2.04,
      "learning_rate": 3.7225794251134646e-05,
      "loss": 0.2488,
      "step": 2704
    },
    {
      "epoch": 2.05,
      "learning_rate": 3.7221066565809385e-05,
      "loss": 0.2844,
      "step": 2705
    },
    {
      "epoch": 2.05,
      "learning_rate": 3.721633888048412e-05,
      "loss": 0.2642,
      "step": 2706
    },
    {
      "epoch": 2.05,
      "learning_rate": 3.721161119515885e-05,
      "loss": 0.2678,
      "step": 2707
    },
    {
      "epoch": 2.05,
      "learning_rate": 3.720688350983359e-05,
      "loss": 0.2248,
      "step": 2708
    },
    {
      "epoch": 2.05,
      "learning_rate": 3.720215582450832e-05,
      "loss": 0.2552,
      "step": 2709
    },
    {
      "epoch": 2.05,
      "learning_rate": 3.719742813918306e-05,
      "loss": 0.2574,
      "step": 2710
    },
    {
      "epoch": 2.05,
      "learning_rate": 3.719270045385779e-05,
      "loss": 0.2593,
      "step": 2711
    },
    {
      "epoch": 2.05,
      "learning_rate": 3.718797276853253e-05,
      "loss": 0.3019,
      "step": 2712
    },
    {
      "epoch": 2.05,
      "learning_rate": 3.718324508320726e-05,
      "loss": 0.2727,
      "step": 2713
    },
    {
      "epoch": 2.05,
      "learning_rate": 3.7178517397882e-05,
      "loss": 0.2565,
      "step": 2714
    },
    {
      "epoch": 2.05,
      "learning_rate": 3.717378971255673e-05,
      "loss": 0.2768,
      "step": 2715
    },
    {
      "epoch": 2.05,
      "learning_rate": 3.716906202723147e-05,
      "loss": 0.2561,
      "step": 2716
    },
    {
      "epoch": 2.05,
      "learning_rate": 3.71643343419062e-05,
      "loss": 0.1999,
      "step": 2717
    },
    {
      "epoch": 2.06,
      "learning_rate": 3.715960665658094e-05,
      "loss": 0.1942,
      "step": 2718
    },
    {
      "epoch": 2.06,
      "learning_rate": 3.7154878971255674e-05,
      "loss": 0.2613,
      "step": 2719
    },
    {
      "epoch": 2.06,
      "learning_rate": 3.715015128593041e-05,
      "loss": 0.2854,
      "step": 2720
    },
    {
      "epoch": 2.06,
      "learning_rate": 3.7145423600605145e-05,
      "loss": 0.2517,
      "step": 2721
    },
    {
      "epoch": 2.06,
      "learning_rate": 3.7140695915279884e-05,
      "loss": 0.2872,
      "step": 2722
    },
    {
      "epoch": 2.06,
      "learning_rate": 3.7135968229954616e-05,
      "loss": 0.2546,
      "step": 2723
    },
    {
      "epoch": 2.06,
      "learning_rate": 3.7131240544629355e-05,
      "loss": 0.2101,
      "step": 2724
    },
    {
      "epoch": 2.06,
      "learning_rate": 3.712651285930409e-05,
      "loss": 0.2735,
      "step": 2725
    },
    {
      "epoch": 2.06,
      "learning_rate": 3.7121785173978826e-05,
      "loss": 0.2575,
      "step": 2726
    },
    {
      "epoch": 2.06,
      "learning_rate": 3.711705748865356e-05,
      "loss": 0.2323,
      "step": 2727
    },
    {
      "epoch": 2.06,
      "learning_rate": 3.71123298033283e-05,
      "loss": 0.3217,
      "step": 2728
    },
    {
      "epoch": 2.06,
      "learning_rate": 3.710760211800303e-05,
      "loss": 0.28,
      "step": 2729
    },
    {
      "epoch": 2.06,
      "learning_rate": 3.710287443267777e-05,
      "loss": 0.256,
      "step": 2730
    },
    {
      "epoch": 2.07,
      "learning_rate": 3.70981467473525e-05,
      "loss": 0.2208,
      "step": 2731
    },
    {
      "epoch": 2.07,
      "learning_rate": 3.709341906202723e-05,
      "loss": 0.2698,
      "step": 2732
    },
    {
      "epoch": 2.07,
      "learning_rate": 3.708869137670197e-05,
      "loss": 0.244,
      "step": 2733
    },
    {
      "epoch": 2.07,
      "learning_rate": 3.70839636913767e-05,
      "loss": 0.2184,
      "step": 2734
    },
    {
      "epoch": 2.07,
      "learning_rate": 3.7079236006051435e-05,
      "loss": 0.2843,
      "step": 2735
    },
    {
      "epoch": 2.07,
      "learning_rate": 3.7074508320726173e-05,
      "loss": 0.2907,
      "step": 2736
    },
    {
      "epoch": 2.07,
      "learning_rate": 3.7069780635400905e-05,
      "loss": 0.2353,
      "step": 2737
    },
    {
      "epoch": 2.07,
      "learning_rate": 3.7065052950075644e-05,
      "loss": 0.2434,
      "step": 2738
    },
    {
      "epoch": 2.07,
      "learning_rate": 3.7060325264750376e-05,
      "loss": 0.2586,
      "step": 2739
    },
    {
      "epoch": 2.07,
      "learning_rate": 3.7055597579425115e-05,
      "loss": 0.2578,
      "step": 2740
    },
    {
      "epoch": 2.07,
      "learning_rate": 3.705086989409985e-05,
      "loss": 0.2875,
      "step": 2741
    },
    {
      "epoch": 2.07,
      "learning_rate": 3.7046142208774586e-05,
      "loss": 0.2484,
      "step": 2742
    },
    {
      "epoch": 2.07,
      "learning_rate": 3.704141452344932e-05,
      "loss": 0.2456,
      "step": 2743
    },
    {
      "epoch": 2.08,
      "learning_rate": 3.703668683812406e-05,
      "loss": 0.2544,
      "step": 2744
    },
    {
      "epoch": 2.08,
      "learning_rate": 3.703195915279879e-05,
      "loss": 0.2664,
      "step": 2745
    },
    {
      "epoch": 2.08,
      "learning_rate": 3.702723146747353e-05,
      "loss": 0.2514,
      "step": 2746
    },
    {
      "epoch": 2.08,
      "learning_rate": 3.702250378214826e-05,
      "loss": 0.2577,
      "step": 2747
    },
    {
      "epoch": 2.08,
      "learning_rate": 3.7017776096823e-05,
      "loss": 0.2732,
      "step": 2748
    },
    {
      "epoch": 2.08,
      "learning_rate": 3.701304841149773e-05,
      "loss": 0.283,
      "step": 2749
    },
    {
      "epoch": 2.08,
      "learning_rate": 3.700832072617247e-05,
      "loss": 0.2741,
      "step": 2750
    },
    {
      "epoch": 2.08,
      "learning_rate": 3.70035930408472e-05,
      "loss": 0.2542,
      "step": 2751
    },
    {
      "epoch": 2.08,
      "learning_rate": 3.699886535552194e-05,
      "loss": 0.275,
      "step": 2752
    },
    {
      "epoch": 2.08,
      "learning_rate": 3.699413767019667e-05,
      "loss": 0.2492,
      "step": 2753
    },
    {
      "epoch": 2.08,
      "learning_rate": 3.698940998487141e-05,
      "loss": 0.2255,
      "step": 2754
    },
    {
      "epoch": 2.08,
      "learning_rate": 3.6984682299546144e-05,
      "loss": 0.2332,
      "step": 2755
    },
    {
      "epoch": 2.08,
      "learning_rate": 3.697995461422088e-05,
      "loss": 0.3261,
      "step": 2756
    },
    {
      "epoch": 2.09,
      "learning_rate": 3.6975226928895614e-05,
      "loss": 0.2309,
      "step": 2757
    },
    {
      "epoch": 2.09,
      "learning_rate": 3.697049924357035e-05,
      "loss": 0.2845,
      "step": 2758
    },
    {
      "epoch": 2.09,
      "learning_rate": 3.6965771558245085e-05,
      "loss": 0.2816,
      "step": 2759
    },
    {
      "epoch": 2.09,
      "learning_rate": 3.696104387291982e-05,
      "loss": 0.2781,
      "step": 2760
    },
    {
      "epoch": 2.09,
      "learning_rate": 3.6956316187594556e-05,
      "loss": 0.2943,
      "step": 2761
    },
    {
      "epoch": 2.09,
      "learning_rate": 3.695158850226929e-05,
      "loss": 0.2698,
      "step": 2762
    },
    {
      "epoch": 2.09,
      "learning_rate": 3.694686081694402e-05,
      "loss": 0.271,
      "step": 2763
    },
    {
      "epoch": 2.09,
      "learning_rate": 3.694213313161876e-05,
      "loss": 0.302,
      "step": 2764
    },
    {
      "epoch": 2.09,
      "learning_rate": 3.693740544629349e-05,
      "loss": 0.3105,
      "step": 2765
    },
    {
      "epoch": 2.09,
      "learning_rate": 3.693267776096823e-05,
      "loss": 0.2694,
      "step": 2766
    },
    {
      "epoch": 2.09,
      "learning_rate": 3.692795007564297e-05,
      "loss": 0.2445,
      "step": 2767
    },
    {
      "epoch": 2.09,
      "learning_rate": 3.69232223903177e-05,
      "loss": 0.2371,
      "step": 2768
    },
    {
      "epoch": 2.09,
      "learning_rate": 3.691849470499244e-05,
      "loss": 0.2385,
      "step": 2769
    },
    {
      "epoch": 2.09,
      "learning_rate": 3.691376701966717e-05,
      "loss": 0.3081,
      "step": 2770
    },
    {
      "epoch": 2.1,
      "learning_rate": 3.690903933434191e-05,
      "loss": 0.245,
      "step": 2771
    },
    {
      "epoch": 2.1,
      "learning_rate": 3.690431164901664e-05,
      "loss": 0.2843,
      "step": 2772
    },
    {
      "epoch": 2.1,
      "learning_rate": 3.689958396369138e-05,
      "loss": 0.2426,
      "step": 2773
    },
    {
      "epoch": 2.1,
      "learning_rate": 3.6894856278366114e-05,
      "loss": 0.2427,
      "step": 2774
    },
    {
      "epoch": 2.1,
      "learning_rate": 3.689012859304085e-05,
      "loss": 0.2574,
      "step": 2775
    },
    {
      "epoch": 2.1,
      "learning_rate": 3.6885400907715585e-05,
      "loss": 0.2348,
      "step": 2776
    },
    {
      "epoch": 2.1,
      "learning_rate": 3.688067322239032e-05,
      "loss": 0.2487,
      "step": 2777
    },
    {
      "epoch": 2.1,
      "learning_rate": 3.6875945537065055e-05,
      "loss": 0.2651,
      "step": 2778
    },
    {
      "epoch": 2.1,
      "learning_rate": 3.6871217851739794e-05,
      "loss": 0.261,
      "step": 2779
    },
    {
      "epoch": 2.1,
      "learning_rate": 3.6866490166414526e-05,
      "loss": 0.2659,
      "step": 2780
    },
    {
      "epoch": 2.1,
      "learning_rate": 3.6861762481089265e-05,
      "loss": 0.2524,
      "step": 2781
    },
    {
      "epoch": 2.1,
      "learning_rate": 3.6857034795764e-05,
      "loss": 0.2175,
      "step": 2782
    },
    {
      "epoch": 2.1,
      "learning_rate": 3.6852307110438736e-05,
      "loss": 0.2337,
      "step": 2783
    },
    {
      "epoch": 2.11,
      "learning_rate": 3.684757942511347e-05,
      "loss": 0.2429,
      "step": 2784
    },
    {
      "epoch": 2.11,
      "learning_rate": 3.68428517397882e-05,
      "loss": 0.2785,
      "step": 2785
    },
    {
      "epoch": 2.11,
      "learning_rate": 3.683812405446294e-05,
      "loss": 0.239,
      "step": 2786
    },
    {
      "epoch": 2.11,
      "learning_rate": 3.683339636913767e-05,
      "loss": 0.2044,
      "step": 2787
    },
    {
      "epoch": 2.11,
      "learning_rate": 3.68286686838124e-05,
      "loss": 0.2619,
      "step": 2788
    },
    {
      "epoch": 2.11,
      "learning_rate": 3.682394099848714e-05,
      "loss": 0.2904,
      "step": 2789
    },
    {
      "epoch": 2.11,
      "learning_rate": 3.6819213313161874e-05,
      "loss": 0.2389,
      "step": 2790
    },
    {
      "epoch": 2.11,
      "learning_rate": 3.681448562783661e-05,
      "loss": 0.2592,
      "step": 2791
    },
    {
      "epoch": 2.11,
      "learning_rate": 3.6809757942511345e-05,
      "loss": 0.239,
      "step": 2792
    },
    {
      "epoch": 2.11,
      "learning_rate": 3.6805030257186084e-05,
      "loss": 0.2952,
      "step": 2793
    },
    {
      "epoch": 2.11,
      "learning_rate": 3.6800302571860816e-05,
      "loss": 0.2828,
      "step": 2794
    },
    {
      "epoch": 2.11,
      "learning_rate": 3.6795574886535555e-05,
      "loss": 0.2629,
      "step": 2795
    },
    {
      "epoch": 2.11,
      "learning_rate": 3.679084720121029e-05,
      "loss": 0.1872,
      "step": 2796
    },
    {
      "epoch": 2.12,
      "learning_rate": 3.6786119515885026e-05,
      "loss": 0.2885,
      "step": 2797
    },
    {
      "epoch": 2.12,
      "learning_rate": 3.678139183055976e-05,
      "loss": 0.2457,
      "step": 2798
    },
    {
      "epoch": 2.12,
      "learning_rate": 3.6776664145234496e-05,
      "loss": 0.2956,
      "step": 2799
    },
    {
      "epoch": 2.12,
      "learning_rate": 3.677193645990923e-05,
      "loss": 0.312,
      "step": 2800
    },
    {
      "epoch": 2.12,
      "learning_rate": 3.676720877458397e-05,
      "loss": 0.2742,
      "step": 2801
    },
    {
      "epoch": 2.12,
      "learning_rate": 3.67624810892587e-05,
      "loss": 0.2343,
      "step": 2802
    },
    {
      "epoch": 2.12,
      "learning_rate": 3.675775340393344e-05,
      "loss": 0.2495,
      "step": 2803
    },
    {
      "epoch": 2.12,
      "learning_rate": 3.675302571860817e-05,
      "loss": 0.257,
      "step": 2804
    },
    {
      "epoch": 2.12,
      "learning_rate": 3.674829803328291e-05,
      "loss": 0.2251,
      "step": 2805
    },
    {
      "epoch": 2.12,
      "learning_rate": 3.674357034795764e-05,
      "loss": 0.2134,
      "step": 2806
    },
    {
      "epoch": 2.12,
      "learning_rate": 3.673884266263238e-05,
      "loss": 0.2624,
      "step": 2807
    },
    {
      "epoch": 2.12,
      "learning_rate": 3.673411497730711e-05,
      "loss": 0.2779,
      "step": 2808
    },
    {
      "epoch": 2.12,
      "learning_rate": 3.672938729198185e-05,
      "loss": 0.2446,
      "step": 2809
    },
    {
      "epoch": 2.13,
      "learning_rate": 3.672465960665658e-05,
      "loss": 0.2504,
      "step": 2810
    },
    {
      "epoch": 2.13,
      "learning_rate": 3.671993192133132e-05,
      "loss": 0.2592,
      "step": 2811
    },
    {
      "epoch": 2.13,
      "learning_rate": 3.6715204236006054e-05,
      "loss": 0.2447,
      "step": 2812
    },
    {
      "epoch": 2.13,
      "learning_rate": 3.6710476550680786e-05,
      "loss": 0.2626,
      "step": 2813
    },
    {
      "epoch": 2.13,
      "learning_rate": 3.6705748865355525e-05,
      "loss": 0.279,
      "step": 2814
    },
    {
      "epoch": 2.13,
      "learning_rate": 3.670102118003026e-05,
      "loss": 0.2851,
      "step": 2815
    },
    {
      "epoch": 2.13,
      "learning_rate": 3.669629349470499e-05,
      "loss": 0.2726,
      "step": 2816
    },
    {
      "epoch": 2.13,
      "learning_rate": 3.669156580937973e-05,
      "loss": 0.2581,
      "step": 2817
    },
    {
      "epoch": 2.13,
      "learning_rate": 3.668683812405446e-05,
      "loss": 0.2982,
      "step": 2818
    },
    {
      "epoch": 2.13,
      "learning_rate": 3.66821104387292e-05,
      "loss": 0.3168,
      "step": 2819
    },
    {
      "epoch": 2.13,
      "learning_rate": 3.667738275340393e-05,
      "loss": 0.2944,
      "step": 2820
    },
    {
      "epoch": 2.13,
      "learning_rate": 3.667265506807867e-05,
      "loss": 0.2579,
      "step": 2821
    },
    {
      "epoch": 2.13,
      "learning_rate": 3.66679273827534e-05,
      "loss": 0.2299,
      "step": 2822
    },
    {
      "epoch": 2.13,
      "learning_rate": 3.666319969742814e-05,
      "loss": 0.2388,
      "step": 2823
    },
    {
      "epoch": 2.14,
      "learning_rate": 3.665847201210287e-05,
      "loss": 0.2128,
      "step": 2824
    },
    {
      "epoch": 2.14,
      "learning_rate": 3.665374432677761e-05,
      "loss": 0.3137,
      "step": 2825
    },
    {
      "epoch": 2.14,
      "learning_rate": 3.664901664145234e-05,
      "loss": 0.2794,
      "step": 2826
    },
    {
      "epoch": 2.14,
      "learning_rate": 3.664428895612708e-05,
      "loss": 0.2448,
      "step": 2827
    },
    {
      "epoch": 2.14,
      "learning_rate": 3.663956127080182e-05,
      "loss": 0.2626,
      "step": 2828
    },
    {
      "epoch": 2.14,
      "learning_rate": 3.663483358547655e-05,
      "loss": 0.2139,
      "step": 2829
    },
    {
      "epoch": 2.14,
      "learning_rate": 3.663010590015129e-05,
      "loss": 0.2377,
      "step": 2830
    },
    {
      "epoch": 2.14,
      "learning_rate": 3.6625378214826024e-05,
      "loss": 0.2433,
      "step": 2831
    },
    {
      "epoch": 2.14,
      "learning_rate": 3.662065052950076e-05,
      "loss": 0.2593,
      "step": 2832
    },
    {
      "epoch": 2.14,
      "learning_rate": 3.6615922844175495e-05,
      "loss": 0.3356,
      "step": 2833
    },
    {
      "epoch": 2.14,
      "learning_rate": 3.6611195158850234e-05,
      "loss": 0.264,
      "step": 2834
    },
    {
      "epoch": 2.14,
      "learning_rate": 3.6606467473524966e-05,
      "loss": 0.2756,
      "step": 2835
    },
    {
      "epoch": 2.14,
      "learning_rate": 3.6601739788199705e-05,
      "loss": 0.2502,
      "step": 2836
    },
    {
      "epoch": 2.15,
      "learning_rate": 3.659701210287444e-05,
      "loss": 0.2292,
      "step": 2837
    },
    {
      "epoch": 2.15,
      "learning_rate": 3.659228441754917e-05,
      "loss": 0.2893,
      "step": 2838
    },
    {
      "epoch": 2.15,
      "learning_rate": 3.658755673222391e-05,
      "loss": 0.3022,
      "step": 2839
    },
    {
      "epoch": 2.15,
      "learning_rate": 3.658282904689864e-05,
      "loss": 0.2701,
      "step": 2840
    },
    {
      "epoch": 2.15,
      "learning_rate": 3.657810136157337e-05,
      "loss": 0.2187,
      "step": 2841
    },
    {
      "epoch": 2.15,
      "learning_rate": 3.657337367624811e-05,
      "loss": 0.2614,
      "step": 2842
    },
    {
      "epoch": 2.15,
      "learning_rate": 3.656864599092284e-05,
      "loss": 0.2533,
      "step": 2843
    },
    {
      "epoch": 2.15,
      "learning_rate": 3.656391830559758e-05,
      "loss": 0.2332,
      "step": 2844
    },
    {
      "epoch": 2.15,
      "learning_rate": 3.6559190620272313e-05,
      "loss": 0.2459,
      "step": 2845
    },
    {
      "epoch": 2.15,
      "learning_rate": 3.655446293494705e-05,
      "loss": 0.2852,
      "step": 2846
    },
    {
      "epoch": 2.15,
      "learning_rate": 3.6549735249621784e-05,
      "loss": 0.3307,
      "step": 2847
    },
    {
      "epoch": 2.15,
      "learning_rate": 3.654500756429652e-05,
      "loss": 0.2791,
      "step": 2848
    },
    {
      "epoch": 2.15,
      "learning_rate": 3.6540279878971255e-05,
      "loss": 0.2654,
      "step": 2849
    },
    {
      "epoch": 2.16,
      "learning_rate": 3.6535552193645994e-05,
      "loss": 0.229,
      "step": 2850
    },
    {
      "epoch": 2.16,
      "learning_rate": 3.6530824508320726e-05,
      "loss": 0.2797,
      "step": 2851
    },
    {
      "epoch": 2.16,
      "learning_rate": 3.6526096822995465e-05,
      "loss": 0.2379,
      "step": 2852
    },
    {
      "epoch": 2.16,
      "learning_rate": 3.65213691376702e-05,
      "loss": 0.2707,
      "step": 2853
    },
    {
      "epoch": 2.16,
      "learning_rate": 3.6516641452344936e-05,
      "loss": 0.2518,
      "step": 2854
    },
    {
      "epoch": 2.16,
      "learning_rate": 3.651191376701967e-05,
      "loss": 0.2569,
      "step": 2855
    },
    {
      "epoch": 2.16,
      "learning_rate": 3.650718608169441e-05,
      "loss": 0.2755,
      "step": 2856
    },
    {
      "epoch": 2.16,
      "learning_rate": 3.650245839636914e-05,
      "loss": 0.2432,
      "step": 2857
    },
    {
      "epoch": 2.16,
      "learning_rate": 3.649773071104388e-05,
      "loss": 0.2544,
      "step": 2858
    },
    {
      "epoch": 2.16,
      "learning_rate": 3.649300302571861e-05,
      "loss": 0.2838,
      "step": 2859
    },
    {
      "epoch": 2.16,
      "learning_rate": 3.648827534039335e-05,
      "loss": 0.2909,
      "step": 2860
    },
    {
      "epoch": 2.16,
      "learning_rate": 3.648354765506808e-05,
      "loss": 0.2449,
      "step": 2861
    },
    {
      "epoch": 2.16,
      "learning_rate": 3.647881996974282e-05,
      "loss": 0.2407,
      "step": 2862
    },
    {
      "epoch": 2.17,
      "learning_rate": 3.647409228441755e-05,
      "loss": 0.279,
      "step": 2863
    },
    {
      "epoch": 2.17,
      "learning_rate": 3.646936459909229e-05,
      "loss": 0.2757,
      "step": 2864
    },
    {
      "epoch": 2.17,
      "learning_rate": 3.646463691376702e-05,
      "loss": 0.2465,
      "step": 2865
    },
    {
      "epoch": 2.17,
      "learning_rate": 3.6459909228441754e-05,
      "loss": 0.2472,
      "step": 2866
    },
    {
      "epoch": 2.17,
      "learning_rate": 3.645518154311649e-05,
      "loss": 0.2336,
      "step": 2867
    },
    {
      "epoch": 2.17,
      "learning_rate": 3.6450453857791225e-05,
      "loss": 0.2852,
      "step": 2868
    },
    {
      "epoch": 2.17,
      "learning_rate": 3.644572617246596e-05,
      "loss": 0.2621,
      "step": 2869
    },
    {
      "epoch": 2.17,
      "learning_rate": 3.6440998487140696e-05,
      "loss": 0.3297,
      "step": 2870
    },
    {
      "epoch": 2.17,
      "learning_rate": 3.643627080181543e-05,
      "loss": 0.3096,
      "step": 2871
    },
    {
      "epoch": 2.17,
      "learning_rate": 3.643154311649017e-05,
      "loss": 0.2645,
      "step": 2872
    },
    {
      "epoch": 2.17,
      "learning_rate": 3.64268154311649e-05,
      "loss": 0.2349,
      "step": 2873
    },
    {
      "epoch": 2.17,
      "learning_rate": 3.642208774583964e-05,
      "loss": 0.2793,
      "step": 2874
    },
    {
      "epoch": 2.17,
      "learning_rate": 3.641736006051437e-05,
      "loss": 0.2485,
      "step": 2875
    },
    {
      "epoch": 2.18,
      "learning_rate": 3.641263237518911e-05,
      "loss": 0.2494,
      "step": 2876
    },
    {
      "epoch": 2.18,
      "learning_rate": 3.640790468986384e-05,
      "loss": 0.269,
      "step": 2877
    },
    {
      "epoch": 2.18,
      "learning_rate": 3.640317700453858e-05,
      "loss": 0.2717,
      "step": 2878
    },
    {
      "epoch": 2.18,
      "learning_rate": 3.639844931921331e-05,
      "loss": 0.2658,
      "step": 2879
    },
    {
      "epoch": 2.18,
      "learning_rate": 3.639372163388805e-05,
      "loss": 0.2768,
      "step": 2880
    },
    {
      "epoch": 2.18,
      "learning_rate": 3.638899394856278e-05,
      "loss": 0.2559,
      "step": 2881
    },
    {
      "epoch": 2.18,
      "learning_rate": 3.638426626323752e-05,
      "loss": 0.2812,
      "step": 2882
    },
    {
      "epoch": 2.18,
      "learning_rate": 3.6379538577912254e-05,
      "loss": 0.2582,
      "step": 2883
    },
    {
      "epoch": 2.18,
      "learning_rate": 3.637481089258699e-05,
      "loss": 0.3224,
      "step": 2884
    },
    {
      "epoch": 2.18,
      "learning_rate": 3.6370083207261725e-05,
      "loss": 0.236,
      "step": 2885
    },
    {
      "epoch": 2.18,
      "learning_rate": 3.6365355521936463e-05,
      "loss": 0.2514,
      "step": 2886
    },
    {
      "epoch": 2.18,
      "learning_rate": 3.63606278366112e-05,
      "loss": 0.2564,
      "step": 2887
    },
    {
      "epoch": 2.18,
      "learning_rate": 3.6355900151285934e-05,
      "loss": 0.2945,
      "step": 2888
    },
    {
      "epoch": 2.18,
      "learning_rate": 3.635117246596067e-05,
      "loss": 0.2447,
      "step": 2889
    },
    {
      "epoch": 2.19,
      "learning_rate": 3.6346444780635405e-05,
      "loss": 0.2953,
      "step": 2890
    },
    {
      "epoch": 2.19,
      "learning_rate": 3.634171709531014e-05,
      "loss": 0.2647,
      "step": 2891
    },
    {
      "epoch": 2.19,
      "learning_rate": 3.6336989409984876e-05,
      "loss": 0.2629,
      "step": 2892
    },
    {
      "epoch": 2.19,
      "learning_rate": 3.633226172465961e-05,
      "loss": 0.2632,
      "step": 2893
    },
    {
      "epoch": 2.19,
      "learning_rate": 3.632753403933434e-05,
      "loss": 0.2734,
      "step": 2894
    },
    {
      "epoch": 2.19,
      "learning_rate": 3.632280635400908e-05,
      "loss": 0.304,
      "step": 2895
    },
    {
      "epoch": 2.19,
      "learning_rate": 3.631807866868381e-05,
      "loss": 0.2398,
      "step": 2896
    },
    {
      "epoch": 2.19,
      "learning_rate": 3.631335098335855e-05,
      "loss": 0.2429,
      "step": 2897
    },
    {
      "epoch": 2.19,
      "learning_rate": 3.630862329803328e-05,
      "loss": 0.2249,
      "step": 2898
    },
    {
      "epoch": 2.19,
      "learning_rate": 3.630389561270802e-05,
      "loss": 0.292,
      "step": 2899
    },
    {
      "epoch": 2.19,
      "learning_rate": 3.629916792738275e-05,
      "loss": 0.307,
      "step": 2900
    },
    {
      "epoch": 2.19,
      "learning_rate": 3.629444024205749e-05,
      "loss": 0.2732,
      "step": 2901
    },
    {
      "epoch": 2.19,
      "learning_rate": 3.6289712556732224e-05,
      "loss": 0.2733,
      "step": 2902
    },
    {
      "epoch": 2.2,
      "learning_rate": 3.628498487140696e-05,
      "loss": 0.2887,
      "step": 2903
    },
    {
      "epoch": 2.2,
      "learning_rate": 3.6280257186081695e-05,
      "loss": 0.3089,
      "step": 2904
    },
    {
      "epoch": 2.2,
      "learning_rate": 3.6275529500756434e-05,
      "loss": 0.2429,
      "step": 2905
    },
    {
      "epoch": 2.2,
      "learning_rate": 3.6270801815431166e-05,
      "loss": 0.27,
      "step": 2906
    },
    {
      "epoch": 2.2,
      "learning_rate": 3.6266074130105904e-05,
      "loss": 0.3267,
      "step": 2907
    },
    {
      "epoch": 2.2,
      "learning_rate": 3.6261346444780636e-05,
      "loss": 0.2715,
      "step": 2908
    },
    {
      "epoch": 2.2,
      "learning_rate": 3.6256618759455375e-05,
      "loss": 0.2749,
      "step": 2909
    },
    {
      "epoch": 2.2,
      "learning_rate": 3.625189107413011e-05,
      "loss": 0.2495,
      "step": 2910
    },
    {
      "epoch": 2.2,
      "learning_rate": 3.6247163388804846e-05,
      "loss": 0.2498,
      "step": 2911
    },
    {
      "epoch": 2.2,
      "learning_rate": 3.624243570347958e-05,
      "loss": 0.2655,
      "step": 2912
    },
    {
      "epoch": 2.2,
      "learning_rate": 3.623770801815432e-05,
      "loss": 0.3053,
      "step": 2913
    },
    {
      "epoch": 2.2,
      "learning_rate": 3.623298033282905e-05,
      "loss": 0.2555,
      "step": 2914
    },
    {
      "epoch": 2.2,
      "learning_rate": 3.622825264750379e-05,
      "loss": 0.278,
      "step": 2915
    },
    {
      "epoch": 2.21,
      "learning_rate": 3.622352496217852e-05,
      "loss": 0.3009,
      "step": 2916
    },
    {
      "epoch": 2.21,
      "learning_rate": 3.621879727685326e-05,
      "loss": 0.2592,
      "step": 2917
    },
    {
      "epoch": 2.21,
      "learning_rate": 3.621406959152799e-05,
      "loss": 0.2373,
      "step": 2918
    },
    {
      "epoch": 2.21,
      "learning_rate": 3.620934190620272e-05,
      "loss": 0.2418,
      "step": 2919
    },
    {
      "epoch": 2.21,
      "learning_rate": 3.620461422087746e-05,
      "loss": 0.2941,
      "step": 2920
    },
    {
      "epoch": 2.21,
      "learning_rate": 3.6199886535552194e-05,
      "loss": 0.236,
      "step": 2921
    },
    {
      "epoch": 2.21,
      "learning_rate": 3.6195158850226926e-05,
      "loss": 0.2764,
      "step": 2922
    },
    {
      "epoch": 2.21,
      "learning_rate": 3.6190431164901665e-05,
      "loss": 0.2967,
      "step": 2923
    },
    {
      "epoch": 2.21,
      "learning_rate": 3.61857034795764e-05,
      "loss": 0.2175,
      "step": 2924
    },
    {
      "epoch": 2.21,
      "learning_rate": 3.6180975794251136e-05,
      "loss": 0.2735,
      "step": 2925
    },
    {
      "epoch": 2.21,
      "learning_rate": 3.617624810892587e-05,
      "loss": 0.3096,
      "step": 2926
    },
    {
      "epoch": 2.21,
      "learning_rate": 3.6171520423600607e-05,
      "loss": 0.2541,
      "step": 2927
    },
    {
      "epoch": 2.21,
      "learning_rate": 3.616679273827534e-05,
      "loss": 0.2466,
      "step": 2928
    },
    {
      "epoch": 2.22,
      "learning_rate": 3.616206505295008e-05,
      "loss": 0.2807,
      "step": 2929
    },
    {
      "epoch": 2.22,
      "learning_rate": 3.615733736762481e-05,
      "loss": 0.2752,
      "step": 2930
    },
    {
      "epoch": 2.22,
      "learning_rate": 3.615260968229955e-05,
      "loss": 0.2675,
      "step": 2931
    },
    {
      "epoch": 2.22,
      "learning_rate": 3.614788199697428e-05,
      "loss": 0.2945,
      "step": 2932
    },
    {
      "epoch": 2.22,
      "learning_rate": 3.614315431164902e-05,
      "loss": 0.2714,
      "step": 2933
    },
    {
      "epoch": 2.22,
      "learning_rate": 3.613842662632375e-05,
      "loss": 0.3129,
      "step": 2934
    },
    {
      "epoch": 2.22,
      "learning_rate": 3.613369894099849e-05,
      "loss": 0.2161,
      "step": 2935
    },
    {
      "epoch": 2.22,
      "learning_rate": 3.612897125567322e-05,
      "loss": 0.2718,
      "step": 2936
    },
    {
      "epoch": 2.22,
      "learning_rate": 3.612424357034796e-05,
      "loss": 0.2491,
      "step": 2937
    },
    {
      "epoch": 2.22,
      "learning_rate": 3.611951588502269e-05,
      "loss": 0.2761,
      "step": 2938
    },
    {
      "epoch": 2.22,
      "learning_rate": 3.611478819969743e-05,
      "loss": 0.3039,
      "step": 2939
    },
    {
      "epoch": 2.22,
      "learning_rate": 3.6110060514372164e-05,
      "loss": 0.3056,
      "step": 2940
    },
    {
      "epoch": 2.22,
      "learning_rate": 3.61053328290469e-05,
      "loss": 0.2628,
      "step": 2941
    },
    {
      "epoch": 2.22,
      "learning_rate": 3.6100605143721635e-05,
      "loss": 0.2657,
      "step": 2942
    },
    {
      "epoch": 2.23,
      "learning_rate": 3.6095877458396374e-05,
      "loss": 0.2787,
      "step": 2943
    },
    {
      "epoch": 2.23,
      "learning_rate": 3.6091149773071106e-05,
      "loss": 0.2741,
      "step": 2944
    },
    {
      "epoch": 2.23,
      "learning_rate": 3.6086422087745845e-05,
      "loss": 0.2263,
      "step": 2945
    },
    {
      "epoch": 2.23,
      "learning_rate": 3.608169440242058e-05,
      "loss": 0.2229,
      "step": 2946
    },
    {
      "epoch": 2.23,
      "learning_rate": 3.607696671709531e-05,
      "loss": 0.3325,
      "step": 2947
    },
    {
      "epoch": 2.23,
      "learning_rate": 3.607223903177005e-05,
      "loss": 0.2832,
      "step": 2948
    },
    {
      "epoch": 2.23,
      "learning_rate": 3.606751134644478e-05,
      "loss": 0.2684,
      "step": 2949
    },
    {
      "epoch": 2.23,
      "learning_rate": 3.606278366111952e-05,
      "loss": 0.2694,
      "step": 2950
    },
    {
      "epoch": 2.23,
      "learning_rate": 3.605805597579425e-05,
      "loss": 0.251,
      "step": 2951
    },
    {
      "epoch": 2.23,
      "learning_rate": 3.605332829046899e-05,
      "loss": 0.2743,
      "step": 2952
    },
    {
      "epoch": 2.23,
      "learning_rate": 3.604860060514372e-05,
      "loss": 0.2717,
      "step": 2953
    },
    {
      "epoch": 2.23,
      "learning_rate": 3.604387291981846e-05,
      "loss": 0.2337,
      "step": 2954
    },
    {
      "epoch": 2.23,
      "learning_rate": 3.603914523449319e-05,
      "loss": 0.2692,
      "step": 2955
    },
    {
      "epoch": 2.24,
      "learning_rate": 3.603441754916793e-05,
      "loss": 0.2429,
      "step": 2956
    },
    {
      "epoch": 2.24,
      "learning_rate": 3.602968986384266e-05,
      "loss": 0.2307,
      "step": 2957
    },
    {
      "epoch": 2.24,
      "learning_rate": 3.60249621785174e-05,
      "loss": 0.3274,
      "step": 2958
    },
    {
      "epoch": 2.24,
      "learning_rate": 3.6020234493192134e-05,
      "loss": 0.2627,
      "step": 2959
    },
    {
      "epoch": 2.24,
      "learning_rate": 3.601550680786687e-05,
      "loss": 0.2751,
      "step": 2960
    },
    {
      "epoch": 2.24,
      "learning_rate": 3.6010779122541605e-05,
      "loss": 0.2745,
      "step": 2961
    },
    {
      "epoch": 2.24,
      "learning_rate": 3.6006051437216344e-05,
      "loss": 0.2634,
      "step": 2962
    },
    {
      "epoch": 2.24,
      "learning_rate": 3.6001323751891076e-05,
      "loss": 0.2853,
      "step": 2963
    },
    {
      "epoch": 2.24,
      "learning_rate": 3.5996596066565815e-05,
      "loss": 0.3089,
      "step": 2964
    },
    {
      "epoch": 2.24,
      "learning_rate": 3.599186838124055e-05,
      "loss": 0.2745,
      "step": 2965
    },
    {
      "epoch": 2.24,
      "learning_rate": 3.5987140695915286e-05,
      "loss": 0.2461,
      "step": 2966
    },
    {
      "epoch": 2.24,
      "learning_rate": 3.598241301059002e-05,
      "loss": 0.2355,
      "step": 2967
    },
    {
      "epoch": 2.24,
      "learning_rate": 3.5977685325264757e-05,
      "loss": 0.272,
      "step": 2968
    },
    {
      "epoch": 2.25,
      "learning_rate": 3.597295763993949e-05,
      "loss": 0.2651,
      "step": 2969
    },
    {
      "epoch": 2.25,
      "learning_rate": 3.596822995461423e-05,
      "loss": 0.2179,
      "step": 2970
    },
    {
      "epoch": 2.25,
      "learning_rate": 3.596350226928896e-05,
      "loss": 0.2707,
      "step": 2971
    },
    {
      "epoch": 2.25,
      "learning_rate": 3.595877458396369e-05,
      "loss": 0.2562,
      "step": 2972
    },
    {
      "epoch": 2.25,
      "learning_rate": 3.595404689863843e-05,
      "loss": 0.3022,
      "step": 2973
    },
    {
      "epoch": 2.25,
      "learning_rate": 3.594931921331316e-05,
      "loss": 0.2911,
      "step": 2974
    },
    {
      "epoch": 2.25,
      "learning_rate": 3.5944591527987895e-05,
      "loss": 0.2958,
      "step": 2975
    },
    {
      "epoch": 2.25,
      "learning_rate": 3.593986384266263e-05,
      "loss": 0.2821,
      "step": 2976
    },
    {
      "epoch": 2.25,
      "learning_rate": 3.5935136157337365e-05,
      "loss": 0.28,
      "step": 2977
    },
    {
      "epoch": 2.25,
      "learning_rate": 3.5930408472012104e-05,
      "loss": 0.2929,
      "step": 2978
    },
    {
      "epoch": 2.25,
      "learning_rate": 3.5925680786686836e-05,
      "loss": 0.2377,
      "step": 2979
    },
    {
      "epoch": 2.25,
      "learning_rate": 3.5920953101361575e-05,
      "loss": 0.2388,
      "step": 2980
    },
    {
      "epoch": 2.25,
      "learning_rate": 3.591622541603631e-05,
      "loss": 0.3139,
      "step": 2981
    },
    {
      "epoch": 2.26,
      "learning_rate": 3.5911497730711046e-05,
      "loss": 0.2913,
      "step": 2982
    },
    {
      "epoch": 2.26,
      "learning_rate": 3.590677004538578e-05,
      "loss": 0.2123,
      "step": 2983
    },
    {
      "epoch": 2.26,
      "learning_rate": 3.590204236006052e-05,
      "loss": 0.2189,
      "step": 2984
    },
    {
      "epoch": 2.26,
      "learning_rate": 3.589731467473525e-05,
      "loss": 0.2883,
      "step": 2985
    },
    {
      "epoch": 2.26,
      "learning_rate": 3.589258698940999e-05,
      "loss": 0.2752,
      "step": 2986
    },
    {
      "epoch": 2.26,
      "learning_rate": 3.588785930408472e-05,
      "loss": 0.2626,
      "step": 2987
    },
    {
      "epoch": 2.26,
      "learning_rate": 3.588313161875946e-05,
      "loss": 0.3082,
      "step": 2988
    },
    {
      "epoch": 2.26,
      "learning_rate": 3.587840393343419e-05,
      "loss": 0.2411,
      "step": 2989
    },
    {
      "epoch": 2.26,
      "learning_rate": 3.587367624810893e-05,
      "loss": 0.3206,
      "step": 2990
    },
    {
      "epoch": 2.26,
      "learning_rate": 3.586894856278366e-05,
      "loss": 0.2397,
      "step": 2991
    },
    {
      "epoch": 2.26,
      "learning_rate": 3.58642208774584e-05,
      "loss": 0.3057,
      "step": 2992
    },
    {
      "epoch": 2.26,
      "learning_rate": 3.585949319213313e-05,
      "loss": 0.295,
      "step": 2993
    },
    {
      "epoch": 2.26,
      "learning_rate": 3.585476550680787e-05,
      "loss": 0.2302,
      "step": 2994
    },
    {
      "epoch": 2.27,
      "learning_rate": 3.5850037821482603e-05,
      "loss": 0.2655,
      "step": 2995
    },
    {
      "epoch": 2.27,
      "learning_rate": 3.584531013615734e-05,
      "loss": 0.262,
      "step": 2996
    },
    {
      "epoch": 2.27,
      "learning_rate": 3.5840582450832074e-05,
      "loss": 0.2157,
      "step": 2997
    },
    {
      "epoch": 2.27,
      "learning_rate": 3.583585476550681e-05,
      "loss": 0.2324,
      "step": 2998
    },
    {
      "epoch": 2.27,
      "learning_rate": 3.5831127080181545e-05,
      "loss": 0.2633,
      "step": 2999
    },
    {
      "epoch": 2.27,
      "learning_rate": 3.582639939485628e-05,
      "loss": 0.2324,
      "step": 3000
    },
    {
      "epoch": 2.27,
      "learning_rate": 3.5821671709531016e-05,
      "loss": 0.2983,
      "step": 3001
    },
    {
      "epoch": 2.27,
      "learning_rate": 3.581694402420575e-05,
      "loss": 0.2683,
      "step": 3002
    },
    {
      "epoch": 2.27,
      "learning_rate": 3.581221633888048e-05,
      "loss": 0.3642,
      "step": 3003
    },
    {
      "epoch": 2.27,
      "learning_rate": 3.580748865355522e-05,
      "loss": 0.2594,
      "step": 3004
    },
    {
      "epoch": 2.27,
      "learning_rate": 3.580276096822995e-05,
      "loss": 0.2576,
      "step": 3005
    },
    {
      "epoch": 2.27,
      "learning_rate": 3.579803328290469e-05,
      "loss": 0.2872,
      "step": 3006
    },
    {
      "epoch": 2.27,
      "learning_rate": 3.579330559757943e-05,
      "loss": 0.272,
      "step": 3007
    },
    {
      "epoch": 2.27,
      "learning_rate": 3.578857791225416e-05,
      "loss": 0.3169,
      "step": 3008
    },
    {
      "epoch": 2.28,
      "learning_rate": 3.57838502269289e-05,
      "loss": 0.2688,
      "step": 3009
    },
    {
      "epoch": 2.28,
      "learning_rate": 3.577912254160363e-05,
      "loss": 0.3303,
      "step": 3010
    },
    {
      "epoch": 2.28,
      "learning_rate": 3.577439485627837e-05,
      "loss": 0.2623,
      "step": 3011
    },
    {
      "epoch": 2.28,
      "learning_rate": 3.57696671709531e-05,
      "loss": 0.238,
      "step": 3012
    },
    {
      "epoch": 2.28,
      "learning_rate": 3.576493948562784e-05,
      "loss": 0.2411,
      "step": 3013
    },
    {
      "epoch": 2.28,
      "learning_rate": 3.5760211800302574e-05,
      "loss": 0.2757,
      "step": 3014
    },
    {
      "epoch": 2.28,
      "learning_rate": 3.575548411497731e-05,
      "loss": 0.3047,
      "step": 3015
    },
    {
      "epoch": 2.28,
      "learning_rate": 3.5750756429652044e-05,
      "loss": 0.3214,
      "step": 3016
    },
    {
      "epoch": 2.28,
      "learning_rate": 3.574602874432678e-05,
      "loss": 0.2689,
      "step": 3017
    },
    {
      "epoch": 2.28,
      "learning_rate": 3.5741301059001515e-05,
      "loss": 0.2482,
      "step": 3018
    },
    {
      "epoch": 2.28,
      "learning_rate": 3.5736573373676254e-05,
      "loss": 0.2524,
      "step": 3019
    },
    {
      "epoch": 2.28,
      "learning_rate": 3.5731845688350986e-05,
      "loss": 0.2939,
      "step": 3020
    },
    {
      "epoch": 2.28,
      "learning_rate": 3.5727118003025725e-05,
      "loss": 0.2464,
      "step": 3021
    },
    {
      "epoch": 2.29,
      "learning_rate": 3.572239031770046e-05,
      "loss": 0.2464,
      "step": 3022
    },
    {
      "epoch": 2.29,
      "learning_rate": 3.5717662632375196e-05,
      "loss": 0.2923,
      "step": 3023
    },
    {
      "epoch": 2.29,
      "learning_rate": 3.571293494704993e-05,
      "loss": 0.3078,
      "step": 3024
    },
    {
      "epoch": 2.29,
      "learning_rate": 3.570820726172466e-05,
      "loss": 0.2647,
      "step": 3025
    },
    {
      "epoch": 2.29,
      "learning_rate": 3.57034795763994e-05,
      "loss": 0.2622,
      "step": 3026
    },
    {
      "epoch": 2.29,
      "learning_rate": 3.569875189107413e-05,
      "loss": 0.216,
      "step": 3027
    },
    {
      "epoch": 2.29,
      "learning_rate": 3.569402420574886e-05,
      "loss": 0.336,
      "step": 3028
    },
    {
      "epoch": 2.29,
      "learning_rate": 3.56892965204236e-05,
      "loss": 0.281,
      "step": 3029
    },
    {
      "epoch": 2.29,
      "learning_rate": 3.5684568835098334e-05,
      "loss": 0.261,
      "step": 3030
    },
    {
      "epoch": 2.29,
      "learning_rate": 3.567984114977307e-05,
      "loss": 0.2385,
      "step": 3031
    },
    {
      "epoch": 2.29,
      "learning_rate": 3.5675113464447805e-05,
      "loss": 0.2734,
      "step": 3032
    },
    {
      "epoch": 2.29,
      "learning_rate": 3.5670385779122544e-05,
      "loss": 0.2842,
      "step": 3033
    },
    {
      "epoch": 2.29,
      "learning_rate": 3.5665658093797276e-05,
      "loss": 0.2832,
      "step": 3034
    },
    {
      "epoch": 2.3,
      "learning_rate": 3.5660930408472015e-05,
      "loss": 0.2715,
      "step": 3035
    },
    {
      "epoch": 2.3,
      "learning_rate": 3.565620272314675e-05,
      "loss": 0.2558,
      "step": 3036
    },
    {
      "epoch": 2.3,
      "learning_rate": 3.5651475037821485e-05,
      "loss": 0.2653,
      "step": 3037
    },
    {
      "epoch": 2.3,
      "learning_rate": 3.564674735249622e-05,
      "loss": 0.2806,
      "step": 3038
    },
    {
      "epoch": 2.3,
      "learning_rate": 3.5642019667170956e-05,
      "loss": 0.2866,
      "step": 3039
    },
    {
      "epoch": 2.3,
      "learning_rate": 3.563729198184569e-05,
      "loss": 0.2525,
      "step": 3040
    },
    {
      "epoch": 2.3,
      "learning_rate": 3.563256429652043e-05,
      "loss": 0.2912,
      "step": 3041
    },
    {
      "epoch": 2.3,
      "learning_rate": 3.562783661119516e-05,
      "loss": 0.2895,
      "step": 3042
    },
    {
      "epoch": 2.3,
      "learning_rate": 3.56231089258699e-05,
      "loss": 0.2923,
      "step": 3043
    },
    {
      "epoch": 2.3,
      "learning_rate": 3.561838124054463e-05,
      "loss": 0.2506,
      "step": 3044
    },
    {
      "epoch": 2.3,
      "learning_rate": 3.561365355521937e-05,
      "loss": 0.2767,
      "step": 3045
    },
    {
      "epoch": 2.3,
      "learning_rate": 3.56089258698941e-05,
      "loss": 0.2881,
      "step": 3046
    },
    {
      "epoch": 2.3,
      "learning_rate": 3.560419818456884e-05,
      "loss": 0.3086,
      "step": 3047
    },
    {
      "epoch": 2.31,
      "learning_rate": 3.559947049924357e-05,
      "loss": 0.2891,
      "step": 3048
    },
    {
      "epoch": 2.31,
      "learning_rate": 3.559474281391831e-05,
      "loss": 0.24,
      "step": 3049
    },
    {
      "epoch": 2.31,
      "learning_rate": 3.559001512859304e-05,
      "loss": 0.4076,
      "step": 3050
    },
    {
      "epoch": 2.31,
      "learning_rate": 3.558528744326778e-05,
      "loss": 0.309,
      "step": 3051
    },
    {
      "epoch": 2.31,
      "learning_rate": 3.5580559757942514e-05,
      "loss": 0.2599,
      "step": 3052
    },
    {
      "epoch": 2.31,
      "learning_rate": 3.5575832072617246e-05,
      "loss": 0.2222,
      "step": 3053
    },
    {
      "epoch": 2.31,
      "learning_rate": 3.5571104387291985e-05,
      "loss": 0.2518,
      "step": 3054
    },
    {
      "epoch": 2.31,
      "learning_rate": 3.556637670196672e-05,
      "loss": 0.2579,
      "step": 3055
    },
    {
      "epoch": 2.31,
      "learning_rate": 3.556164901664145e-05,
      "loss": 0.3135,
      "step": 3056
    },
    {
      "epoch": 2.31,
      "learning_rate": 3.555692133131619e-05,
      "loss": 0.2626,
      "step": 3057
    },
    {
      "epoch": 2.31,
      "learning_rate": 3.555219364599092e-05,
      "loss": 0.3095,
      "step": 3058
    },
    {
      "epoch": 2.31,
      "learning_rate": 3.554746596066566e-05,
      "loss": 0.2544,
      "step": 3059
    },
    {
      "epoch": 2.31,
      "learning_rate": 3.554273827534039e-05,
      "loss": 0.2281,
      "step": 3060
    },
    {
      "epoch": 2.31,
      "learning_rate": 3.553801059001513e-05,
      "loss": 0.2653,
      "step": 3061
    },
    {
      "epoch": 2.32,
      "learning_rate": 3.553328290468986e-05,
      "loss": 0.2777,
      "step": 3062
    },
    {
      "epoch": 2.32,
      "learning_rate": 3.55285552193646e-05,
      "loss": 0.2558,
      "step": 3063
    },
    {
      "epoch": 2.32,
      "learning_rate": 3.552382753403933e-05,
      "loss": 0.3037,
      "step": 3064
    },
    {
      "epoch": 2.32,
      "learning_rate": 3.551909984871407e-05,
      "loss": 0.2546,
      "step": 3065
    },
    {
      "epoch": 2.32,
      "learning_rate": 3.55143721633888e-05,
      "loss": 0.2805,
      "step": 3066
    },
    {
      "epoch": 2.32,
      "learning_rate": 3.550964447806354e-05,
      "loss": 0.333,
      "step": 3067
    },
    {
      "epoch": 2.32,
      "learning_rate": 3.550491679273828e-05,
      "loss": 0.2777,
      "step": 3068
    },
    {
      "epoch": 2.32,
      "learning_rate": 3.550018910741301e-05,
      "loss": 0.3305,
      "step": 3069
    },
    {
      "epoch": 2.32,
      "learning_rate": 3.549546142208775e-05,
      "loss": 0.2771,
      "step": 3070
    },
    {
      "epoch": 2.32,
      "learning_rate": 3.5490733736762484e-05,
      "loss": 0.2773,
      "step": 3071
    },
    {
      "epoch": 2.32,
      "learning_rate": 3.548600605143722e-05,
      "loss": 0.252,
      "step": 3072
    },
    {
      "epoch": 2.32,
      "learning_rate": 3.5481278366111955e-05,
      "loss": 0.2395,
      "step": 3073
    },
    {
      "epoch": 2.32,
      "learning_rate": 3.5476550680786694e-05,
      "loss": 0.3641,
      "step": 3074
    },
    {
      "epoch": 2.33,
      "learning_rate": 3.5471822995461426e-05,
      "loss": 0.2856,
      "step": 3075
    },
    {
      "epoch": 2.33,
      "learning_rate": 3.5467095310136165e-05,
      "loss": 0.2664,
      "step": 3076
    },
    {
      "epoch": 2.33,
      "learning_rate": 3.5462367624810897e-05,
      "loss": 0.2223,
      "step": 3077
    },
    {
      "epoch": 2.33,
      "learning_rate": 3.545763993948563e-05,
      "loss": 0.2937,
      "step": 3078
    },
    {
      "epoch": 2.33,
      "learning_rate": 3.545291225416037e-05,
      "loss": 0.2876,
      "step": 3079
    },
    {
      "epoch": 2.33,
      "learning_rate": 3.54481845688351e-05,
      "loss": 0.2765,
      "step": 3080
    },
    {
      "epoch": 2.33,
      "learning_rate": 3.544345688350983e-05,
      "loss": 0.3236,
      "step": 3081
    },
    {
      "epoch": 2.33,
      "learning_rate": 3.543872919818457e-05,
      "loss": 0.2807,
      "step": 3082
    },
    {
      "epoch": 2.33,
      "learning_rate": 3.54340015128593e-05,
      "loss": 0.2539,
      "step": 3083
    },
    {
      "epoch": 2.33,
      "learning_rate": 3.542927382753404e-05,
      "loss": 0.2709,
      "step": 3084
    },
    {
      "epoch": 2.33,
      "learning_rate": 3.5424546142208773e-05,
      "loss": 0.2461,
      "step": 3085
    },
    {
      "epoch": 2.33,
      "learning_rate": 3.541981845688351e-05,
      "loss": 0.2945,
      "step": 3086
    },
    {
      "epoch": 2.33,
      "learning_rate": 3.5415090771558244e-05,
      "loss": 0.2368,
      "step": 3087
    },
    {
      "epoch": 2.34,
      "learning_rate": 3.541036308623298e-05,
      "loss": 0.2564,
      "step": 3088
    },
    {
      "epoch": 2.34,
      "learning_rate": 3.5405635400907715e-05,
      "loss": 0.262,
      "step": 3089
    },
    {
      "epoch": 2.34,
      "learning_rate": 3.5400907715582454e-05,
      "loss": 0.2886,
      "step": 3090
    },
    {
      "epoch": 2.34,
      "learning_rate": 3.5396180030257186e-05,
      "loss": 0.2591,
      "step": 3091
    },
    {
      "epoch": 2.34,
      "learning_rate": 3.5391452344931925e-05,
      "loss": 0.2613,
      "step": 3092
    },
    {
      "epoch": 2.34,
      "learning_rate": 3.538672465960666e-05,
      "loss": 0.2816,
      "step": 3093
    },
    {
      "epoch": 2.34,
      "learning_rate": 3.5381996974281396e-05,
      "loss": 0.343,
      "step": 3094
    },
    {
      "epoch": 2.34,
      "learning_rate": 3.537726928895613e-05,
      "loss": 0.2496,
      "step": 3095
    },
    {
      "epoch": 2.34,
      "learning_rate": 3.537254160363087e-05,
      "loss": 0.277,
      "step": 3096
    },
    {
      "epoch": 2.34,
      "learning_rate": 3.53678139183056e-05,
      "loss": 0.2814,
      "step": 3097
    },
    {
      "epoch": 2.34,
      "learning_rate": 3.536308623298034e-05,
      "loss": 0.2779,
      "step": 3098
    },
    {
      "epoch": 2.34,
      "learning_rate": 3.535835854765507e-05,
      "loss": 0.2858,
      "step": 3099
    },
    {
      "epoch": 2.34,
      "learning_rate": 3.535363086232981e-05,
      "loss": 0.2825,
      "step": 3100
    },
    {
      "epoch": 2.35,
      "learning_rate": 3.534890317700454e-05,
      "loss": 0.3011,
      "step": 3101
    },
    {
      "epoch": 2.35,
      "learning_rate": 3.534417549167928e-05,
      "loss": 0.2669,
      "step": 3102
    },
    {
      "epoch": 2.35,
      "learning_rate": 3.533944780635401e-05,
      "loss": 0.2876,
      "step": 3103
    },
    {
      "epoch": 2.35,
      "learning_rate": 3.533472012102875e-05,
      "loss": 0.2549,
      "step": 3104
    },
    {
      "epoch": 2.35,
      "learning_rate": 3.532999243570348e-05,
      "loss": 0.2593,
      "step": 3105
    },
    {
      "epoch": 2.35,
      "learning_rate": 3.5325264750378214e-05,
      "loss": 0.2641,
      "step": 3106
    },
    {
      "epoch": 2.35,
      "learning_rate": 3.532053706505295e-05,
      "loss": 0.2464,
      "step": 3107
    },
    {
      "epoch": 2.35,
      "learning_rate": 3.5315809379727685e-05,
      "loss": 0.2546,
      "step": 3108
    },
    {
      "epoch": 2.35,
      "learning_rate": 3.531108169440242e-05,
      "loss": 0.2633,
      "step": 3109
    },
    {
      "epoch": 2.35,
      "learning_rate": 3.5306354009077156e-05,
      "loss": 0.2978,
      "step": 3110
    },
    {
      "epoch": 2.35,
      "learning_rate": 3.530162632375189e-05,
      "loss": 0.2233,
      "step": 3111
    },
    {
      "epoch": 2.35,
      "learning_rate": 3.529689863842663e-05,
      "loss": 0.3115,
      "step": 3112
    },
    {
      "epoch": 2.35,
      "learning_rate": 3.529217095310136e-05,
      "loss": 0.2668,
      "step": 3113
    },
    {
      "epoch": 2.36,
      "learning_rate": 3.52874432677761e-05,
      "loss": 0.2648,
      "step": 3114
    },
    {
      "epoch": 2.36,
      "learning_rate": 3.528271558245083e-05,
      "loss": 0.2671,
      "step": 3115
    },
    {
      "epoch": 2.36,
      "learning_rate": 3.527798789712557e-05,
      "loss": 0.3128,
      "step": 3116
    },
    {
      "epoch": 2.36,
      "learning_rate": 3.52732602118003e-05,
      "loss": 0.2942,
      "step": 3117
    },
    {
      "epoch": 2.36,
      "learning_rate": 3.526853252647504e-05,
      "loss": 0.2396,
      "step": 3118
    },
    {
      "epoch": 2.36,
      "learning_rate": 3.526380484114977e-05,
      "loss": 0.2912,
      "step": 3119
    },
    {
      "epoch": 2.36,
      "learning_rate": 3.525907715582451e-05,
      "loss": 0.2679,
      "step": 3120
    },
    {
      "epoch": 2.36,
      "learning_rate": 3.525434947049924e-05,
      "loss": 0.3025,
      "step": 3121
    },
    {
      "epoch": 2.36,
      "learning_rate": 3.524962178517398e-05,
      "loss": 0.2542,
      "step": 3122
    },
    {
      "epoch": 2.36,
      "learning_rate": 3.5244894099848714e-05,
      "loss": 0.2765,
      "step": 3123
    },
    {
      "epoch": 2.36,
      "learning_rate": 3.524016641452345e-05,
      "loss": 0.259,
      "step": 3124
    },
    {
      "epoch": 2.36,
      "learning_rate": 3.5235438729198185e-05,
      "loss": 0.232,
      "step": 3125
    },
    {
      "epoch": 2.36,
      "learning_rate": 3.523071104387292e-05,
      "loss": 0.2403,
      "step": 3126
    },
    {
      "epoch": 2.36,
      "learning_rate": 3.522598335854766e-05,
      "loss": 0.2456,
      "step": 3127
    },
    {
      "epoch": 2.37,
      "learning_rate": 3.5221255673222394e-05,
      "loss": 0.2664,
      "step": 3128
    },
    {
      "epoch": 2.37,
      "learning_rate": 3.521652798789713e-05,
      "loss": 0.2911,
      "step": 3129
    },
    {
      "epoch": 2.37,
      "learning_rate": 3.5211800302571865e-05,
      "loss": 0.27,
      "step": 3130
    },
    {
      "epoch": 2.37,
      "learning_rate": 3.52070726172466e-05,
      "loss": 0.3046,
      "step": 3131
    },
    {
      "epoch": 2.37,
      "learning_rate": 3.5202344931921336e-05,
      "loss": 0.2677,
      "step": 3132
    },
    {
      "epoch": 2.37,
      "learning_rate": 3.519761724659607e-05,
      "loss": 0.2951,
      "step": 3133
    },
    {
      "epoch": 2.37,
      "learning_rate": 3.51928895612708e-05,
      "loss": 0.2486,
      "step": 3134
    },
    {
      "epoch": 2.37,
      "learning_rate": 3.518816187594554e-05,
      "loss": 0.29,
      "step": 3135
    },
    {
      "epoch": 2.37,
      "learning_rate": 3.518343419062027e-05,
      "loss": 0.2739,
      "step": 3136
    },
    {
      "epoch": 2.37,
      "learning_rate": 3.517870650529501e-05,
      "loss": 0.254,
      "step": 3137
    },
    {
      "epoch": 2.37,
      "learning_rate": 3.517397881996974e-05,
      "loss": 0.3151,
      "step": 3138
    },
    {
      "epoch": 2.37,
      "learning_rate": 3.516925113464448e-05,
      "loss": 0.2726,
      "step": 3139
    },
    {
      "epoch": 2.37,
      "learning_rate": 3.516452344931921e-05,
      "loss": 0.2619,
      "step": 3140
    },
    {
      "epoch": 2.38,
      "learning_rate": 3.515979576399395e-05,
      "loss": 0.2496,
      "step": 3141
    },
    {
      "epoch": 2.38,
      "learning_rate": 3.5155068078668684e-05,
      "loss": 0.2564,
      "step": 3142
    },
    {
      "epoch": 2.38,
      "learning_rate": 3.515034039334342e-05,
      "loss": 0.2466,
      "step": 3143
    },
    {
      "epoch": 2.38,
      "learning_rate": 3.5145612708018155e-05,
      "loss": 0.2581,
      "step": 3144
    },
    {
      "epoch": 2.38,
      "learning_rate": 3.5140885022692893e-05,
      "loss": 0.255,
      "step": 3145
    },
    {
      "epoch": 2.38,
      "learning_rate": 3.5136157337367626e-05,
      "loss": 0.224,
      "step": 3146
    },
    {
      "epoch": 2.38,
      "learning_rate": 3.5131429652042364e-05,
      "loss": 0.2965,
      "step": 3147
    },
    {
      "epoch": 2.38,
      "learning_rate": 3.5126701966717096e-05,
      "loss": 0.28,
      "step": 3148
    },
    {
      "epoch": 2.38,
      "learning_rate": 3.5121974281391835e-05,
      "loss": 0.2324,
      "step": 3149
    },
    {
      "epoch": 2.38,
      "learning_rate": 3.511724659606657e-05,
      "loss": 0.2589,
      "step": 3150
    },
    {
      "epoch": 2.38,
      "learning_rate": 3.5112518910741306e-05,
      "loss": 0.2474,
      "step": 3151
    },
    {
      "epoch": 2.38,
      "learning_rate": 3.510779122541604e-05,
      "loss": 0.2788,
      "step": 3152
    },
    {
      "epoch": 2.38,
      "learning_rate": 3.510306354009078e-05,
      "loss": 0.2377,
      "step": 3153
    },
    {
      "epoch": 2.39,
      "learning_rate": 3.509833585476551e-05,
      "loss": 0.2978,
      "step": 3154
    },
    {
      "epoch": 2.39,
      "learning_rate": 3.509360816944025e-05,
      "loss": 0.2691,
      "step": 3155
    },
    {
      "epoch": 2.39,
      "learning_rate": 3.508888048411498e-05,
      "loss": 0.2944,
      "step": 3156
    },
    {
      "epoch": 2.39,
      "learning_rate": 3.508415279878972e-05,
      "loss": 0.2483,
      "step": 3157
    },
    {
      "epoch": 2.39,
      "learning_rate": 3.507942511346445e-05,
      "loss": 0.2723,
      "step": 3158
    },
    {
      "epoch": 2.39,
      "learning_rate": 3.507469742813918e-05,
      "loss": 0.2184,
      "step": 3159
    },
    {
      "epoch": 2.39,
      "learning_rate": 3.506996974281392e-05,
      "loss": 0.2679,
      "step": 3160
    },
    {
      "epoch": 2.39,
      "learning_rate": 3.5065242057488654e-05,
      "loss": 0.3087,
      "step": 3161
    },
    {
      "epoch": 2.39,
      "learning_rate": 3.5060514372163386e-05,
      "loss": 0.2938,
      "step": 3162
    },
    {
      "epoch": 2.39,
      "learning_rate": 3.5055786686838125e-05,
      "loss": 0.2586,
      "step": 3163
    },
    {
      "epoch": 2.39,
      "learning_rate": 3.505105900151286e-05,
      "loss": 0.2672,
      "step": 3164
    },
    {
      "epoch": 2.39,
      "learning_rate": 3.5046331316187596e-05,
      "loss": 0.2712,
      "step": 3165
    },
    {
      "epoch": 2.39,
      "learning_rate": 3.504160363086233e-05,
      "loss": 0.2151,
      "step": 3166
    },
    {
      "epoch": 2.4,
      "learning_rate": 3.5036875945537067e-05,
      "loss": 0.2905,
      "step": 3167
    },
    {
      "epoch": 2.4,
      "learning_rate": 3.50321482602118e-05,
      "loss": 0.275,
      "step": 3168
    },
    {
      "epoch": 2.4,
      "learning_rate": 3.502742057488654e-05,
      "loss": 0.2317,
      "step": 3169
    },
    {
      "epoch": 2.4,
      "learning_rate": 3.502269288956127e-05,
      "loss": 0.2669,
      "step": 3170
    },
    {
      "epoch": 2.4,
      "learning_rate": 3.501796520423601e-05,
      "loss": 0.3015,
      "step": 3171
    },
    {
      "epoch": 2.4,
      "learning_rate": 3.501323751891074e-05,
      "loss": 0.2528,
      "step": 3172
    },
    {
      "epoch": 2.4,
      "learning_rate": 3.500850983358548e-05,
      "loss": 0.2416,
      "step": 3173
    },
    {
      "epoch": 2.4,
      "learning_rate": 3.500378214826021e-05,
      "loss": 0.283,
      "step": 3174
    },
    {
      "epoch": 2.4,
      "learning_rate": 3.499905446293495e-05,
      "loss": 0.2143,
      "step": 3175
    },
    {
      "epoch": 2.4,
      "learning_rate": 3.499432677760968e-05,
      "loss": 0.3041,
      "step": 3176
    },
    {
      "epoch": 2.4,
      "learning_rate": 3.498959909228442e-05,
      "loss": 0.2494,
      "step": 3177
    },
    {
      "epoch": 2.4,
      "learning_rate": 3.498487140695915e-05,
      "loss": 0.2402,
      "step": 3178
    },
    {
      "epoch": 2.4,
      "learning_rate": 3.498014372163389e-05,
      "loss": 0.2632,
      "step": 3179
    },
    {
      "epoch": 2.4,
      "learning_rate": 3.4975416036308624e-05,
      "loss": 0.2742,
      "step": 3180
    },
    {
      "epoch": 2.41,
      "learning_rate": 3.497068835098336e-05,
      "loss": 0.2999,
      "step": 3181
    },
    {
      "epoch": 2.41,
      "learning_rate": 3.4965960665658095e-05,
      "loss": 0.2552,
      "step": 3182
    },
    {
      "epoch": 2.41,
      "learning_rate": 3.4961232980332834e-05,
      "loss": 0.2507,
      "step": 3183
    },
    {
      "epoch": 2.41,
      "learning_rate": 3.4956505295007566e-05,
      "loss": 0.2823,
      "step": 3184
    },
    {
      "epoch": 2.41,
      "learning_rate": 3.4951777609682305e-05,
      "loss": 0.2219,
      "step": 3185
    },
    {
      "epoch": 2.41,
      "learning_rate": 3.494704992435704e-05,
      "loss": 0.2443,
      "step": 3186
    },
    {
      "epoch": 2.41,
      "learning_rate": 3.494232223903177e-05,
      "loss": 0.2283,
      "step": 3187
    },
    {
      "epoch": 2.41,
      "learning_rate": 3.493759455370651e-05,
      "loss": 0.3149,
      "step": 3188
    },
    {
      "epoch": 2.41,
      "learning_rate": 3.493286686838124e-05,
      "loss": 0.2912,
      "step": 3189
    },
    {
      "epoch": 2.41,
      "learning_rate": 3.492813918305598e-05,
      "loss": 0.2653,
      "step": 3190
    },
    {
      "epoch": 2.41,
      "learning_rate": 3.492341149773071e-05,
      "loss": 0.2822,
      "step": 3191
    },
    {
      "epoch": 2.41,
      "learning_rate": 3.491868381240545e-05,
      "loss": 0.2957,
      "step": 3192
    },
    {
      "epoch": 2.41,
      "learning_rate": 3.491395612708018e-05,
      "loss": 0.2609,
      "step": 3193
    },
    {
      "epoch": 2.42,
      "learning_rate": 3.490922844175492e-05,
      "loss": 0.2672,
      "step": 3194
    },
    {
      "epoch": 2.42,
      "learning_rate": 3.490450075642965e-05,
      "loss": 0.2896,
      "step": 3195
    },
    {
      "epoch": 2.42,
      "learning_rate": 3.489977307110439e-05,
      "loss": 0.312,
      "step": 3196
    },
    {
      "epoch": 2.42,
      "learning_rate": 3.489504538577912e-05,
      "loss": 0.2666,
      "step": 3197
    },
    {
      "epoch": 2.42,
      "learning_rate": 3.489031770045386e-05,
      "loss": 0.3118,
      "step": 3198
    },
    {
      "epoch": 2.42,
      "learning_rate": 3.4885590015128594e-05,
      "loss": 0.2617,
      "step": 3199
    },
    {
      "epoch": 2.42,
      "learning_rate": 3.488086232980333e-05,
      "loss": 0.2823,
      "step": 3200
    },
    {
      "epoch": 2.42,
      "learning_rate": 3.4876134644478065e-05,
      "loss": 0.3045,
      "step": 3201
    },
    {
      "epoch": 2.42,
      "learning_rate": 3.4871406959152804e-05,
      "loss": 0.2436,
      "step": 3202
    },
    {
      "epoch": 2.42,
      "learning_rate": 3.4866679273827536e-05,
      "loss": 0.2789,
      "step": 3203
    },
    {
      "epoch": 2.42,
      "learning_rate": 3.4861951588502275e-05,
      "loss": 0.272,
      "step": 3204
    },
    {
      "epoch": 2.42,
      "learning_rate": 3.485722390317701e-05,
      "loss": 0.2825,
      "step": 3205
    },
    {
      "epoch": 2.42,
      "learning_rate": 3.4852496217851746e-05,
      "loss": 0.306,
      "step": 3206
    },
    {
      "epoch": 2.43,
      "learning_rate": 3.484776853252648e-05,
      "loss": 0.2707,
      "step": 3207
    },
    {
      "epoch": 2.43,
      "learning_rate": 3.4843040847201216e-05,
      "loss": 0.2678,
      "step": 3208
    },
    {
      "epoch": 2.43,
      "learning_rate": 3.483831316187595e-05,
      "loss": 0.2416,
      "step": 3209
    },
    {
      "epoch": 2.43,
      "learning_rate": 3.483358547655069e-05,
      "loss": 0.2872,
      "step": 3210
    },
    {
      "epoch": 2.43,
      "learning_rate": 3.482885779122542e-05,
      "loss": 0.3126,
      "step": 3211
    },
    {
      "epoch": 2.43,
      "learning_rate": 3.482413010590015e-05,
      "loss": 0.2593,
      "step": 3212
    },
    {
      "epoch": 2.43,
      "learning_rate": 3.481940242057489e-05,
      "loss": 0.2992,
      "step": 3213
    },
    {
      "epoch": 2.43,
      "learning_rate": 3.481467473524962e-05,
      "loss": 0.2775,
      "step": 3214
    },
    {
      "epoch": 2.43,
      "learning_rate": 3.4809947049924354e-05,
      "loss": 0.2501,
      "step": 3215
    },
    {
      "epoch": 2.43,
      "learning_rate": 3.480521936459909e-05,
      "loss": 0.2738,
      "step": 3216
    },
    {
      "epoch": 2.43,
      "learning_rate": 3.4800491679273825e-05,
      "loss": 0.2975,
      "step": 3217
    },
    {
      "epoch": 2.43,
      "learning_rate": 3.4795763993948564e-05,
      "loss": 0.347,
      "step": 3218
    },
    {
      "epoch": 2.43,
      "learning_rate": 3.4791036308623296e-05,
      "loss": 0.2766,
      "step": 3219
    },
    {
      "epoch": 2.44,
      "learning_rate": 3.4786308623298035e-05,
      "loss": 0.291,
      "step": 3220
    },
    {
      "epoch": 2.44,
      "learning_rate": 3.478158093797277e-05,
      "loss": 0.2856,
      "step": 3221
    },
    {
      "epoch": 2.44,
      "learning_rate": 3.4776853252647506e-05,
      "loss": 0.2321,
      "step": 3222
    },
    {
      "epoch": 2.44,
      "learning_rate": 3.477212556732224e-05,
      "loss": 0.2915,
      "step": 3223
    },
    {
      "epoch": 2.44,
      "learning_rate": 3.476739788199698e-05,
      "loss": 0.2923,
      "step": 3224
    },
    {
      "epoch": 2.44,
      "learning_rate": 3.476267019667171e-05,
      "loss": 0.2787,
      "step": 3225
    },
    {
      "epoch": 2.44,
      "learning_rate": 3.475794251134645e-05,
      "loss": 0.2497,
      "step": 3226
    },
    {
      "epoch": 2.44,
      "learning_rate": 3.475321482602118e-05,
      "loss": 0.256,
      "step": 3227
    },
    {
      "epoch": 2.44,
      "learning_rate": 3.474848714069592e-05,
      "loss": 0.3188,
      "step": 3228
    },
    {
      "epoch": 2.44,
      "learning_rate": 3.474375945537065e-05,
      "loss": 0.2701,
      "step": 3229
    },
    {
      "epoch": 2.44,
      "learning_rate": 3.473903177004539e-05,
      "loss": 0.2759,
      "step": 3230
    },
    {
      "epoch": 2.44,
      "learning_rate": 3.473430408472012e-05,
      "loss": 0.3698,
      "step": 3231
    },
    {
      "epoch": 2.44,
      "learning_rate": 3.472957639939486e-05,
      "loss": 0.3248,
      "step": 3232
    },
    {
      "epoch": 2.45,
      "learning_rate": 3.472484871406959e-05,
      "loss": 0.2402,
      "step": 3233
    },
    {
      "epoch": 2.45,
      "learning_rate": 3.472012102874433e-05,
      "loss": 0.2618,
      "step": 3234
    },
    {
      "epoch": 2.45,
      "learning_rate": 3.4715393343419063e-05,
      "loss": 0.2464,
      "step": 3235
    },
    {
      "epoch": 2.45,
      "learning_rate": 3.47106656580938e-05,
      "loss": 0.2388,
      "step": 3236
    },
    {
      "epoch": 2.45,
      "learning_rate": 3.4705937972768534e-05,
      "loss": 0.2734,
      "step": 3237
    },
    {
      "epoch": 2.45,
      "learning_rate": 3.470121028744327e-05,
      "loss": 0.2867,
      "step": 3238
    },
    {
      "epoch": 2.45,
      "learning_rate": 3.4696482602118005e-05,
      "loss": 0.2892,
      "step": 3239
    },
    {
      "epoch": 2.45,
      "learning_rate": 3.469175491679274e-05,
      "loss": 0.265,
      "step": 3240
    },
    {
      "epoch": 2.45,
      "learning_rate": 3.468702723146747e-05,
      "loss": 0.3099,
      "step": 3241
    },
    {
      "epoch": 2.45,
      "learning_rate": 3.468229954614221e-05,
      "loss": 0.2561,
      "step": 3242
    },
    {
      "epoch": 2.45,
      "learning_rate": 3.467757186081694e-05,
      "loss": 0.2723,
      "step": 3243
    },
    {
      "epoch": 2.45,
      "learning_rate": 3.467284417549168e-05,
      "loss": 0.281,
      "step": 3244
    },
    {
      "epoch": 2.45,
      "learning_rate": 3.466811649016641e-05,
      "loss": 0.259,
      "step": 3245
    },
    {
      "epoch": 2.45,
      "learning_rate": 3.466338880484115e-05,
      "loss": 0.2697,
      "step": 3246
    },
    {
      "epoch": 2.46,
      "learning_rate": 3.465866111951589e-05,
      "loss": 0.2638,
      "step": 3247
    },
    {
      "epoch": 2.46,
      "learning_rate": 3.465393343419062e-05,
      "loss": 0.2919,
      "step": 3248
    },
    {
      "epoch": 2.46,
      "learning_rate": 3.464920574886536e-05,
      "loss": 0.2492,
      "step": 3249
    },
    {
      "epoch": 2.46,
      "learning_rate": 3.464447806354009e-05,
      "loss": 0.265,
      "step": 3250
    },
    {
      "epoch": 2.46,
      "learning_rate": 3.463975037821483e-05,
      "loss": 0.2666,
      "step": 3251
    },
    {
      "epoch": 2.46,
      "learning_rate": 3.463502269288956e-05,
      "loss": 0.2554,
      "step": 3252
    },
    {
      "epoch": 2.46,
      "learning_rate": 3.46302950075643e-05,
      "loss": 0.2571,
      "step": 3253
    },
    {
      "epoch": 2.46,
      "learning_rate": 3.4625567322239034e-05,
      "loss": 0.2694,
      "step": 3254
    },
    {
      "epoch": 2.46,
      "learning_rate": 3.462083963691377e-05,
      "loss": 0.2611,
      "step": 3255
    },
    {
      "epoch": 2.46,
      "learning_rate": 3.4616111951588504e-05,
      "loss": 0.2932,
      "step": 3256
    },
    {
      "epoch": 2.46,
      "learning_rate": 3.461138426626324e-05,
      "loss": 0.2593,
      "step": 3257
    },
    {
      "epoch": 2.46,
      "learning_rate": 3.4606656580937975e-05,
      "loss": 0.2723,
      "step": 3258
    },
    {
      "epoch": 2.46,
      "learning_rate": 3.4601928895612714e-05,
      "loss": 0.2643,
      "step": 3259
    },
    {
      "epoch": 2.47,
      "learning_rate": 3.4597201210287446e-05,
      "loss": 0.2502,
      "step": 3260
    },
    {
      "epoch": 2.47,
      "learning_rate": 3.4592473524962185e-05,
      "loss": 0.2476,
      "step": 3261
    },
    {
      "epoch": 2.47,
      "learning_rate": 3.458774583963692e-05,
      "loss": 0.316,
      "step": 3262
    },
    {
      "epoch": 2.47,
      "learning_rate": 3.4583018154311656e-05,
      "loss": 0.2971,
      "step": 3263
    },
    {
      "epoch": 2.47,
      "learning_rate": 3.457829046898639e-05,
      "loss": 0.3174,
      "step": 3264
    },
    {
      "epoch": 2.47,
      "learning_rate": 3.457356278366112e-05,
      "loss": 0.2817,
      "step": 3265
    },
    {
      "epoch": 2.47,
      "learning_rate": 3.456883509833586e-05,
      "loss": 0.2412,
      "step": 3266
    },
    {
      "epoch": 2.47,
      "learning_rate": 3.456410741301059e-05,
      "loss": 0.2779,
      "step": 3267
    },
    {
      "epoch": 2.47,
      "learning_rate": 3.455937972768532e-05,
      "loss": 0.2495,
      "step": 3268
    },
    {
      "epoch": 2.47,
      "learning_rate": 3.455465204236006e-05,
      "loss": 0.2683,
      "step": 3269
    },
    {
      "epoch": 2.47,
      "learning_rate": 3.4549924357034794e-05,
      "loss": 0.3031,
      "step": 3270
    },
    {
      "epoch": 2.47,
      "learning_rate": 3.454519667170953e-05,
      "loss": 0.2926,
      "step": 3271
    },
    {
      "epoch": 2.47,
      "learning_rate": 3.4540468986384265e-05,
      "loss": 0.2624,
      "step": 3272
    },
    {
      "epoch": 2.48,
      "learning_rate": 3.4535741301059004e-05,
      "loss": 0.2834,
      "step": 3273
    },
    {
      "epoch": 2.48,
      "learning_rate": 3.4531013615733736e-05,
      "loss": 0.2689,
      "step": 3274
    },
    {
      "epoch": 2.48,
      "learning_rate": 3.4526285930408475e-05,
      "loss": 0.2758,
      "step": 3275
    },
    {
      "epoch": 2.48,
      "learning_rate": 3.4521558245083207e-05,
      "loss": 0.2835,
      "step": 3276
    },
    {
      "epoch": 2.48,
      "learning_rate": 3.4516830559757945e-05,
      "loss": 0.2713,
      "step": 3277
    },
    {
      "epoch": 2.48,
      "learning_rate": 3.451210287443268e-05,
      "loss": 0.2752,
      "step": 3278
    },
    {
      "epoch": 2.48,
      "learning_rate": 3.4507375189107416e-05,
      "loss": 0.286,
      "step": 3279
    },
    {
      "epoch": 2.48,
      "learning_rate": 3.450264750378215e-05,
      "loss": 0.2983,
      "step": 3280
    },
    {
      "epoch": 2.48,
      "learning_rate": 3.449791981845689e-05,
      "loss": 0.2985,
      "step": 3281
    },
    {
      "epoch": 2.48,
      "learning_rate": 3.449319213313162e-05,
      "loss": 0.2943,
      "step": 3282
    },
    {
      "epoch": 2.48,
      "learning_rate": 3.448846444780636e-05,
      "loss": 0.2576,
      "step": 3283
    },
    {
      "epoch": 2.48,
      "learning_rate": 3.448373676248109e-05,
      "loss": 0.2935,
      "step": 3284
    },
    {
      "epoch": 2.48,
      "learning_rate": 3.447900907715583e-05,
      "loss": 0.2858,
      "step": 3285
    },
    {
      "epoch": 2.49,
      "learning_rate": 3.447428139183056e-05,
      "loss": 0.284,
      "step": 3286
    },
    {
      "epoch": 2.49,
      "learning_rate": 3.44695537065053e-05,
      "loss": 0.2753,
      "step": 3287
    },
    {
      "epoch": 2.49,
      "learning_rate": 3.446482602118003e-05,
      "loss": 0.2627,
      "step": 3288
    },
    {
      "epoch": 2.49,
      "learning_rate": 3.446009833585477e-05,
      "loss": 0.2945,
      "step": 3289
    },
    {
      "epoch": 2.49,
      "learning_rate": 3.44553706505295e-05,
      "loss": 0.3036,
      "step": 3290
    },
    {
      "epoch": 2.49,
      "learning_rate": 3.445064296520424e-05,
      "loss": 0.2683,
      "step": 3291
    },
    {
      "epoch": 2.49,
      "learning_rate": 3.4445915279878974e-05,
      "loss": 0.3225,
      "step": 3292
    },
    {
      "epoch": 2.49,
      "learning_rate": 3.4441187594553706e-05,
      "loss": 0.2688,
      "step": 3293
    },
    {
      "epoch": 2.49,
      "learning_rate": 3.443645990922844e-05,
      "loss": 0.2946,
      "step": 3294
    },
    {
      "epoch": 2.49,
      "learning_rate": 3.443173222390318e-05,
      "loss": 0.2816,
      "step": 3295
    },
    {
      "epoch": 2.49,
      "learning_rate": 3.442700453857791e-05,
      "loss": 0.2735,
      "step": 3296
    },
    {
      "epoch": 2.49,
      "learning_rate": 3.442227685325265e-05,
      "loss": 0.2892,
      "step": 3297
    },
    {
      "epoch": 2.49,
      "learning_rate": 3.441754916792738e-05,
      "loss": 0.2696,
      "step": 3298
    },
    {
      "epoch": 2.49,
      "learning_rate": 3.441282148260212e-05,
      "loss": 0.2933,
      "step": 3299
    },
    {
      "epoch": 2.5,
      "learning_rate": 3.440809379727685e-05,
      "loss": 0.2291,
      "step": 3300
    },
    {
      "epoch": 2.5,
      "learning_rate": 3.440336611195159e-05,
      "loss": 0.2782,
      "step": 3301
    },
    {
      "epoch": 2.5,
      "learning_rate": 3.439863842662632e-05,
      "loss": 0.2499,
      "step": 3302
    },
    {
      "epoch": 2.5,
      "learning_rate": 3.439391074130106e-05,
      "loss": 0.2834,
      "step": 3303
    },
    {
      "epoch": 2.5,
      "learning_rate": 3.438918305597579e-05,
      "loss": 0.2554,
      "step": 3304
    },
    {
      "epoch": 2.5,
      "learning_rate": 3.438445537065053e-05,
      "loss": 0.277,
      "step": 3305
    },
    {
      "epoch": 2.5,
      "learning_rate": 3.437972768532526e-05,
      "loss": 0.2573,
      "step": 3306
    },
    {
      "epoch": 2.5,
      "learning_rate": 3.4375e-05,
      "loss": 0.2466,
      "step": 3307
    },
    {
      "epoch": 2.5,
      "learning_rate": 3.437027231467474e-05,
      "loss": 0.2469,
      "step": 3308
    },
    {
      "epoch": 2.5,
      "learning_rate": 3.436554462934947e-05,
      "loss": 0.2887,
      "step": 3309
    },
    {
      "epoch": 2.5,
      "learning_rate": 3.436081694402421e-05,
      "loss": 0.2373,
      "step": 3310
    },
    {
      "epoch": 2.5,
      "learning_rate": 3.4356089258698944e-05,
      "loss": 0.3025,
      "step": 3311
    },
    {
      "epoch": 2.5,
      "learning_rate": 3.435136157337368e-05,
      "loss": 0.2516,
      "step": 3312
    },
    {
      "epoch": 2.51,
      "learning_rate": 3.4346633888048415e-05,
      "loss": 0.2789,
      "step": 3313
    },
    {
      "epoch": 2.51,
      "learning_rate": 3.4341906202723154e-05,
      "loss": 0.2754,
      "step": 3314
    },
    {
      "epoch": 2.51,
      "learning_rate": 3.4337178517397886e-05,
      "loss": 0.3028,
      "step": 3315
    },
    {
      "epoch": 2.51,
      "learning_rate": 3.4332450832072624e-05,
      "loss": 0.3073,
      "step": 3316
    },
    {
      "epoch": 2.51,
      "learning_rate": 3.4327723146747357e-05,
      "loss": 0.2916,
      "step": 3317
    },
    {
      "epoch": 2.51,
      "learning_rate": 3.432299546142209e-05,
      "loss": 0.2853,
      "step": 3318
    },
    {
      "epoch": 2.51,
      "learning_rate": 3.431826777609683e-05,
      "loss": 0.3111,
      "step": 3319
    },
    {
      "epoch": 2.51,
      "learning_rate": 3.431354009077156e-05,
      "loss": 0.2222,
      "step": 3320
    },
    {
      "epoch": 2.51,
      "learning_rate": 3.430881240544629e-05,
      "loss": 0.2523,
      "step": 3321
    },
    {
      "epoch": 2.51,
      "learning_rate": 3.430408472012103e-05,
      "loss": 0.308,
      "step": 3322
    },
    {
      "epoch": 2.51,
      "learning_rate": 3.429935703479576e-05,
      "loss": 0.3219,
      "step": 3323
    },
    {
      "epoch": 2.51,
      "learning_rate": 3.42946293494705e-05,
      "loss": 0.2615,
      "step": 3324
    },
    {
      "epoch": 2.51,
      "learning_rate": 3.428990166414523e-05,
      "loss": 0.3042,
      "step": 3325
    },
    {
      "epoch": 2.52,
      "learning_rate": 3.428517397881997e-05,
      "loss": 0.2535,
      "step": 3326
    },
    {
      "epoch": 2.52,
      "learning_rate": 3.4280446293494704e-05,
      "loss": 0.2567,
      "step": 3327
    },
    {
      "epoch": 2.52,
      "learning_rate": 3.427571860816944e-05,
      "loss": 0.2694,
      "step": 3328
    },
    {
      "epoch": 2.52,
      "learning_rate": 3.4270990922844175e-05,
      "loss": 0.2644,
      "step": 3329
    },
    {
      "epoch": 2.52,
      "learning_rate": 3.4266263237518914e-05,
      "loss": 0.2861,
      "step": 3330
    },
    {
      "epoch": 2.52,
      "learning_rate": 3.4261535552193646e-05,
      "loss": 0.2723,
      "step": 3331
    },
    {
      "epoch": 2.52,
      "learning_rate": 3.4256807866868385e-05,
      "loss": 0.2501,
      "step": 3332
    },
    {
      "epoch": 2.52,
      "learning_rate": 3.425208018154312e-05,
      "loss": 0.304,
      "step": 3333
    },
    {
      "epoch": 2.52,
      "learning_rate": 3.4247352496217856e-05,
      "loss": 0.2658,
      "step": 3334
    },
    {
      "epoch": 2.52,
      "learning_rate": 3.424262481089259e-05,
      "loss": 0.246,
      "step": 3335
    },
    {
      "epoch": 2.52,
      "learning_rate": 3.423789712556733e-05,
      "loss": 0.3094,
      "step": 3336
    },
    {
      "epoch": 2.52,
      "learning_rate": 3.423316944024206e-05,
      "loss": 0.3336,
      "step": 3337
    },
    {
      "epoch": 2.52,
      "learning_rate": 3.42284417549168e-05,
      "loss": 0.244,
      "step": 3338
    },
    {
      "epoch": 2.53,
      "learning_rate": 3.422371406959153e-05,
      "loss": 0.3024,
      "step": 3339
    },
    {
      "epoch": 2.53,
      "learning_rate": 3.421898638426627e-05,
      "loss": 0.328,
      "step": 3340
    },
    {
      "epoch": 2.53,
      "learning_rate": 3.4214258698941e-05,
      "loss": 0.2616,
      "step": 3341
    },
    {
      "epoch": 2.53,
      "learning_rate": 3.420953101361574e-05,
      "loss": 0.3112,
      "step": 3342
    },
    {
      "epoch": 2.53,
      "learning_rate": 3.420480332829047e-05,
      "loss": 0.2408,
      "step": 3343
    },
    {
      "epoch": 2.53,
      "learning_rate": 3.420007564296521e-05,
      "loss": 0.294,
      "step": 3344
    },
    {
      "epoch": 2.53,
      "learning_rate": 3.419534795763994e-05,
      "loss": 0.2579,
      "step": 3345
    },
    {
      "epoch": 2.53,
      "learning_rate": 3.4190620272314674e-05,
      "loss": 0.2817,
      "step": 3346
    },
    {
      "epoch": 2.53,
      "learning_rate": 3.4185892586989406e-05,
      "loss": 0.2125,
      "step": 3347
    },
    {
      "epoch": 2.53,
      "learning_rate": 3.4181164901664145e-05,
      "loss": 0.2642,
      "step": 3348
    },
    {
      "epoch": 2.53,
      "learning_rate": 3.417643721633888e-05,
      "loss": 0.3077,
      "step": 3349
    },
    {
      "epoch": 2.53,
      "learning_rate": 3.4171709531013616e-05,
      "loss": 0.2644,
      "step": 3350
    },
    {
      "epoch": 2.53,
      "learning_rate": 3.416698184568835e-05,
      "loss": 0.3123,
      "step": 3351
    },
    {
      "epoch": 2.54,
      "learning_rate": 3.416225416036309e-05,
      "loss": 0.3246,
      "step": 3352
    },
    {
      "epoch": 2.54,
      "learning_rate": 3.415752647503782e-05,
      "loss": 0.2681,
      "step": 3353
    },
    {
      "epoch": 2.54,
      "learning_rate": 3.415279878971256e-05,
      "loss": 0.2872,
      "step": 3354
    },
    {
      "epoch": 2.54,
      "learning_rate": 3.414807110438729e-05,
      "loss": 0.3027,
      "step": 3355
    },
    {
      "epoch": 2.54,
      "learning_rate": 3.414334341906203e-05,
      "loss": 0.3095,
      "step": 3356
    },
    {
      "epoch": 2.54,
      "learning_rate": 3.413861573373676e-05,
      "loss": 0.2956,
      "step": 3357
    },
    {
      "epoch": 2.54,
      "learning_rate": 3.41338880484115e-05,
      "loss": 0.2621,
      "step": 3358
    },
    {
      "epoch": 2.54,
      "learning_rate": 3.412916036308623e-05,
      "loss": 0.2879,
      "step": 3359
    },
    {
      "epoch": 2.54,
      "learning_rate": 3.412443267776097e-05,
      "loss": 0.2785,
      "step": 3360
    },
    {
      "epoch": 2.54,
      "learning_rate": 3.41197049924357e-05,
      "loss": 0.2778,
      "step": 3361
    },
    {
      "epoch": 2.54,
      "learning_rate": 3.411497730711044e-05,
      "loss": 0.3125,
      "step": 3362
    },
    {
      "epoch": 2.54,
      "learning_rate": 3.4110249621785174e-05,
      "loss": 0.2511,
      "step": 3363
    },
    {
      "epoch": 2.54,
      "learning_rate": 3.410552193645991e-05,
      "loss": 0.331,
      "step": 3364
    },
    {
      "epoch": 2.54,
      "learning_rate": 3.4100794251134644e-05,
      "loss": 0.2546,
      "step": 3365
    },
    {
      "epoch": 2.55,
      "learning_rate": 3.409606656580938e-05,
      "loss": 0.2808,
      "step": 3366
    },
    {
      "epoch": 2.55,
      "learning_rate": 3.4091338880484115e-05,
      "loss": 0.2685,
      "step": 3367
    },
    {
      "epoch": 2.55,
      "learning_rate": 3.4086611195158854e-05,
      "loss": 0.2997,
      "step": 3368
    },
    {
      "epoch": 2.55,
      "learning_rate": 3.408188350983359e-05,
      "loss": 0.2507,
      "step": 3369
    },
    {
      "epoch": 2.55,
      "learning_rate": 3.4077155824508325e-05,
      "loss": 0.2837,
      "step": 3370
    },
    {
      "epoch": 2.55,
      "learning_rate": 3.407242813918306e-05,
      "loss": 0.3038,
      "step": 3371
    },
    {
      "epoch": 2.55,
      "learning_rate": 3.4067700453857796e-05,
      "loss": 0.2796,
      "step": 3372
    },
    {
      "epoch": 2.55,
      "learning_rate": 3.406297276853253e-05,
      "loss": 0.2732,
      "step": 3373
    },
    {
      "epoch": 2.55,
      "learning_rate": 3.405824508320726e-05,
      "loss": 0.2871,
      "step": 3374
    },
    {
      "epoch": 2.55,
      "learning_rate": 3.4053517397882e-05,
      "loss": 0.2451,
      "step": 3375
    },
    {
      "epoch": 2.55,
      "learning_rate": 3.404878971255673e-05,
      "loss": 0.2573,
      "step": 3376
    },
    {
      "epoch": 2.55,
      "learning_rate": 3.404406202723147e-05,
      "loss": 0.2738,
      "step": 3377
    },
    {
      "epoch": 2.55,
      "learning_rate": 3.40393343419062e-05,
      "loss": 0.2494,
      "step": 3378
    },
    {
      "epoch": 2.56,
      "learning_rate": 3.403460665658094e-05,
      "loss": 0.2651,
      "step": 3379
    },
    {
      "epoch": 2.56,
      "learning_rate": 3.402987897125567e-05,
      "loss": 0.2854,
      "step": 3380
    },
    {
      "epoch": 2.56,
      "learning_rate": 3.402515128593041e-05,
      "loss": 0.3245,
      "step": 3381
    },
    {
      "epoch": 2.56,
      "learning_rate": 3.4020423600605144e-05,
      "loss": 0.2788,
      "step": 3382
    },
    {
      "epoch": 2.56,
      "learning_rate": 3.401569591527988e-05,
      "loss": 0.3214,
      "step": 3383
    },
    {
      "epoch": 2.56,
      "learning_rate": 3.4010968229954615e-05,
      "loss": 0.2972,
      "step": 3384
    },
    {
      "epoch": 2.56,
      "learning_rate": 3.4006240544629353e-05,
      "loss": 0.2658,
      "step": 3385
    },
    {
      "epoch": 2.56,
      "learning_rate": 3.4001512859304085e-05,
      "loss": 0.2198,
      "step": 3386
    },
    {
      "epoch": 2.56,
      "learning_rate": 3.3996785173978824e-05,
      "loss": 0.2855,
      "step": 3387
    },
    {
      "epoch": 2.56,
      "learning_rate": 3.3992057488653556e-05,
      "loss": 0.3037,
      "step": 3388
    },
    {
      "epoch": 2.56,
      "learning_rate": 3.3987329803328295e-05,
      "loss": 0.267,
      "step": 3389
    },
    {
      "epoch": 2.56,
      "learning_rate": 3.398260211800303e-05,
      "loss": 0.2223,
      "step": 3390
    },
    {
      "epoch": 2.56,
      "learning_rate": 3.3977874432677766e-05,
      "loss": 0.2627,
      "step": 3391
    },
    {
      "epoch": 2.57,
      "learning_rate": 3.39731467473525e-05,
      "loss": 0.29,
      "step": 3392
    },
    {
      "epoch": 2.57,
      "learning_rate": 3.396841906202724e-05,
      "loss": 0.2713,
      "step": 3393
    },
    {
      "epoch": 2.57,
      "learning_rate": 3.396369137670197e-05,
      "loss": 0.2535,
      "step": 3394
    },
    {
      "epoch": 2.57,
      "learning_rate": 3.395896369137671e-05,
      "loss": 0.2943,
      "step": 3395
    },
    {
      "epoch": 2.57,
      "learning_rate": 3.395423600605144e-05,
      "loss": 0.2234,
      "step": 3396
    },
    {
      "epoch": 2.57,
      "learning_rate": 3.394950832072618e-05,
      "loss": 0.2728,
      "step": 3397
    },
    {
      "epoch": 2.57,
      "learning_rate": 3.394478063540091e-05,
      "loss": 0.3008,
      "step": 3398
    },
    {
      "epoch": 2.57,
      "learning_rate": 3.394005295007564e-05,
      "loss": 0.2705,
      "step": 3399
    },
    {
      "epoch": 2.57,
      "learning_rate": 3.3935325264750375e-05,
      "loss": 0.3052,
      "step": 3400
    },
    {
      "epoch": 2.57,
      "learning_rate": 3.3930597579425114e-05,
      "loss": 0.414,
      "step": 3401
    },
    {
      "epoch": 2.57,
      "learning_rate": 3.3925869894099846e-05,
      "loss": 0.3519,
      "step": 3402
    },
    {
      "epoch": 2.57,
      "learning_rate": 3.3921142208774585e-05,
      "loss": 0.3175,
      "step": 3403
    },
    {
      "epoch": 2.57,
      "learning_rate": 3.391641452344932e-05,
      "loss": 0.2971,
      "step": 3404
    },
    {
      "epoch": 2.58,
      "learning_rate": 3.3911686838124056e-05,
      "loss": 0.2916,
      "step": 3405
    },
    {
      "epoch": 2.58,
      "learning_rate": 3.390695915279879e-05,
      "loss": 0.2994,
      "step": 3406
    },
    {
      "epoch": 2.58,
      "learning_rate": 3.3902231467473526e-05,
      "loss": 0.2827,
      "step": 3407
    },
    {
      "epoch": 2.58,
      "learning_rate": 3.389750378214826e-05,
      "loss": 0.2559,
      "step": 3408
    },
    {
      "epoch": 2.58,
      "learning_rate": 3.3892776096823e-05,
      "loss": 0.2848,
      "step": 3409
    },
    {
      "epoch": 2.58,
      "learning_rate": 3.388804841149773e-05,
      "loss": 0.2452,
      "step": 3410
    },
    {
      "epoch": 2.58,
      "learning_rate": 3.388332072617247e-05,
      "loss": 0.2652,
      "step": 3411
    },
    {
      "epoch": 2.58,
      "learning_rate": 3.38785930408472e-05,
      "loss": 0.2676,
      "step": 3412
    },
    {
      "epoch": 2.58,
      "learning_rate": 3.387386535552194e-05,
      "loss": 0.2782,
      "step": 3413
    },
    {
      "epoch": 2.58,
      "learning_rate": 3.386913767019667e-05,
      "loss": 0.3443,
      "step": 3414
    },
    {
      "epoch": 2.58,
      "learning_rate": 3.386440998487141e-05,
      "loss": 0.2682,
      "step": 3415
    },
    {
      "epoch": 2.58,
      "learning_rate": 3.385968229954614e-05,
      "loss": 0.3629,
      "step": 3416
    },
    {
      "epoch": 2.58,
      "learning_rate": 3.385495461422088e-05,
      "loss": 0.2678,
      "step": 3417
    },
    {
      "epoch": 2.58,
      "learning_rate": 3.385022692889561e-05,
      "loss": 0.2559,
      "step": 3418
    },
    {
      "epoch": 2.59,
      "learning_rate": 3.384549924357035e-05,
      "loss": 0.2517,
      "step": 3419
    },
    {
      "epoch": 2.59,
      "learning_rate": 3.3840771558245084e-05,
      "loss": 0.2893,
      "step": 3420
    },
    {
      "epoch": 2.59,
      "learning_rate": 3.383604387291982e-05,
      "loss": 0.277,
      "step": 3421
    },
    {
      "epoch": 2.59,
      "learning_rate": 3.3831316187594555e-05,
      "loss": 0.2751,
      "step": 3422
    },
    {
      "epoch": 2.59,
      "learning_rate": 3.3826588502269294e-05,
      "loss": 0.2683,
      "step": 3423
    },
    {
      "epoch": 2.59,
      "learning_rate": 3.3821860816944026e-05,
      "loss": 0.2815,
      "step": 3424
    },
    {
      "epoch": 2.59,
      "learning_rate": 3.3817133131618765e-05,
      "loss": 0.283,
      "step": 3425
    },
    {
      "epoch": 2.59,
      "learning_rate": 3.3812405446293497e-05,
      "loss": 0.2657,
      "step": 3426
    },
    {
      "epoch": 2.59,
      "learning_rate": 3.380767776096823e-05,
      "loss": 0.3034,
      "step": 3427
    },
    {
      "epoch": 2.59,
      "learning_rate": 3.380295007564297e-05,
      "loss": 0.2597,
      "step": 3428
    },
    {
      "epoch": 2.59,
      "learning_rate": 3.37982223903177e-05,
      "loss": 0.2558,
      "step": 3429
    },
    {
      "epoch": 2.59,
      "learning_rate": 3.379349470499244e-05,
      "loss": 0.2983,
      "step": 3430
    },
    {
      "epoch": 2.59,
      "learning_rate": 3.378876701966717e-05,
      "loss": 0.2252,
      "step": 3431
    },
    {
      "epoch": 2.6,
      "learning_rate": 3.378403933434191e-05,
      "loss": 0.2593,
      "step": 3432
    },
    {
      "epoch": 2.6,
      "learning_rate": 3.377931164901664e-05,
      "loss": 0.3102,
      "step": 3433
    },
    {
      "epoch": 2.6,
      "learning_rate": 3.377458396369138e-05,
      "loss": 0.2533,
      "step": 3434
    },
    {
      "epoch": 2.6,
      "learning_rate": 3.376985627836611e-05,
      "loss": 0.2535,
      "step": 3435
    },
    {
      "epoch": 2.6,
      "learning_rate": 3.376512859304085e-05,
      "loss": 0.3076,
      "step": 3436
    },
    {
      "epoch": 2.6,
      "learning_rate": 3.376040090771558e-05,
      "loss": 0.2422,
      "step": 3437
    },
    {
      "epoch": 2.6,
      "learning_rate": 3.375567322239032e-05,
      "loss": 0.2948,
      "step": 3438
    },
    {
      "epoch": 2.6,
      "learning_rate": 3.3750945537065054e-05,
      "loss": 0.2827,
      "step": 3439
    },
    {
      "epoch": 2.6,
      "learning_rate": 3.374621785173979e-05,
      "loss": 0.2469,
      "step": 3440
    },
    {
      "epoch": 2.6,
      "learning_rate": 3.3741490166414525e-05,
      "loss": 0.3106,
      "step": 3441
    },
    {
      "epoch": 2.6,
      "learning_rate": 3.3736762481089264e-05,
      "loss": 0.2891,
      "step": 3442
    },
    {
      "epoch": 2.6,
      "learning_rate": 3.3732034795763996e-05,
      "loss": 0.2545,
      "step": 3443
    },
    {
      "epoch": 2.6,
      "learning_rate": 3.3727307110438735e-05,
      "loss": 0.2963,
      "step": 3444
    },
    {
      "epoch": 2.61,
      "learning_rate": 3.372257942511347e-05,
      "loss": 0.2444,
      "step": 3445
    },
    {
      "epoch": 2.61,
      "learning_rate": 3.3717851739788206e-05,
      "loss": 0.2781,
      "step": 3446
    },
    {
      "epoch": 2.61,
      "learning_rate": 3.371312405446294e-05,
      "loss": 0.3104,
      "step": 3447
    },
    {
      "epoch": 2.61,
      "learning_rate": 3.3708396369137676e-05,
      "loss": 0.2794,
      "step": 3448
    },
    {
      "epoch": 2.61,
      "learning_rate": 3.370366868381241e-05,
      "loss": 0.2886,
      "step": 3449
    },
    {
      "epoch": 2.61,
      "learning_rate": 3.369894099848715e-05,
      "loss": 0.3208,
      "step": 3450
    },
    {
      "epoch": 2.61,
      "learning_rate": 3.369421331316188e-05,
      "loss": 0.2714,
      "step": 3451
    },
    {
      "epoch": 2.61,
      "learning_rate": 3.368948562783661e-05,
      "loss": 0.2596,
      "step": 3452
    },
    {
      "epoch": 2.61,
      "learning_rate": 3.3684757942511344e-05,
      "loss": 0.2287,
      "step": 3453
    },
    {
      "epoch": 2.61,
      "learning_rate": 3.368003025718608e-05,
      "loss": 0.2879,
      "step": 3454
    },
    {
      "epoch": 2.61,
      "learning_rate": 3.3675302571860814e-05,
      "loss": 0.3039,
      "step": 3455
    },
    {
      "epoch": 2.61,
      "learning_rate": 3.367057488653555e-05,
      "loss": 0.2477,
      "step": 3456
    },
    {
      "epoch": 2.61,
      "learning_rate": 3.3665847201210285e-05,
      "loss": 0.3168,
      "step": 3457
    },
    {
      "epoch": 2.62,
      "learning_rate": 3.3661119515885024e-05,
      "loss": 0.2612,
      "step": 3458
    },
    {
      "epoch": 2.62,
      "learning_rate": 3.3656391830559756e-05,
      "loss": 0.297,
      "step": 3459
    },
    {
      "epoch": 2.62,
      "learning_rate": 3.3651664145234495e-05,
      "loss": 0.3138,
      "step": 3460
    },
    {
      "epoch": 2.62,
      "learning_rate": 3.364693645990923e-05,
      "loss": 0.2261,
      "step": 3461
    },
    {
      "epoch": 2.62,
      "learning_rate": 3.3642208774583966e-05,
      "loss": 0.269,
      "step": 3462
    },
    {
      "epoch": 2.62,
      "learning_rate": 3.36374810892587e-05,
      "loss": 0.2798,
      "step": 3463
    },
    {
      "epoch": 2.62,
      "learning_rate": 3.363275340393344e-05,
      "loss": 0.263,
      "step": 3464
    },
    {
      "epoch": 2.62,
      "learning_rate": 3.362802571860817e-05,
      "loss": 0.3136,
      "step": 3465
    },
    {
      "epoch": 2.62,
      "learning_rate": 3.362329803328291e-05,
      "loss": 0.2386,
      "step": 3466
    },
    {
      "epoch": 2.62,
      "learning_rate": 3.361857034795764e-05,
      "loss": 0.2924,
      "step": 3467
    },
    {
      "epoch": 2.62,
      "learning_rate": 3.361384266263238e-05,
      "loss": 0.2349,
      "step": 3468
    },
    {
      "epoch": 2.62,
      "learning_rate": 3.360911497730711e-05,
      "loss": 0.3269,
      "step": 3469
    },
    {
      "epoch": 2.62,
      "learning_rate": 3.360438729198185e-05,
      "loss": 0.3092,
      "step": 3470
    },
    {
      "epoch": 2.63,
      "learning_rate": 3.359965960665658e-05,
      "loss": 0.2714,
      "step": 3471
    },
    {
      "epoch": 2.63,
      "learning_rate": 3.359493192133132e-05,
      "loss": 0.288,
      "step": 3472
    },
    {
      "epoch": 2.63,
      "learning_rate": 3.359020423600605e-05,
      "loss": 0.3036,
      "step": 3473
    },
    {
      "epoch": 2.63,
      "learning_rate": 3.358547655068079e-05,
      "loss": 0.3068,
      "step": 3474
    },
    {
      "epoch": 2.63,
      "learning_rate": 3.358074886535552e-05,
      "loss": 0.3168,
      "step": 3475
    },
    {
      "epoch": 2.63,
      "learning_rate": 3.357602118003026e-05,
      "loss": 0.3102,
      "step": 3476
    },
    {
      "epoch": 2.63,
      "learning_rate": 3.3571293494704994e-05,
      "loss": 0.2878,
      "step": 3477
    },
    {
      "epoch": 2.63,
      "learning_rate": 3.356656580937973e-05,
      "loss": 0.3015,
      "step": 3478
    },
    {
      "epoch": 2.63,
      "learning_rate": 3.3561838124054465e-05,
      "loss": 0.2636,
      "step": 3479
    },
    {
      "epoch": 2.63,
      "learning_rate": 3.35571104387292e-05,
      "loss": 0.2488,
      "step": 3480
    },
    {
      "epoch": 2.63,
      "learning_rate": 3.355238275340393e-05,
      "loss": 0.2755,
      "step": 3481
    },
    {
      "epoch": 2.63,
      "learning_rate": 3.354765506807867e-05,
      "loss": 0.2688,
      "step": 3482
    },
    {
      "epoch": 2.63,
      "learning_rate": 3.35429273827534e-05,
      "loss": 0.2407,
      "step": 3483
    },
    {
      "epoch": 2.63,
      "learning_rate": 3.353819969742814e-05,
      "loss": 0.2291,
      "step": 3484
    },
    {
      "epoch": 2.64,
      "learning_rate": 3.353347201210287e-05,
      "loss": 0.2868,
      "step": 3485
    },
    {
      "epoch": 2.64,
      "learning_rate": 3.352874432677761e-05,
      "loss": 0.2715,
      "step": 3486
    },
    {
      "epoch": 2.64,
      "learning_rate": 3.352401664145234e-05,
      "loss": 0.2491,
      "step": 3487
    },
    {
      "epoch": 2.64,
      "learning_rate": 3.351928895612708e-05,
      "loss": 0.2484,
      "step": 3488
    },
    {
      "epoch": 2.64,
      "learning_rate": 3.351456127080182e-05,
      "loss": 0.3249,
      "step": 3489
    },
    {
      "epoch": 2.64,
      "learning_rate": 3.350983358547655e-05,
      "loss": 0.2606,
      "step": 3490
    },
    {
      "epoch": 2.64,
      "learning_rate": 3.350510590015129e-05,
      "loss": 0.2559,
      "step": 3491
    },
    {
      "epoch": 2.64,
      "learning_rate": 3.350037821482602e-05,
      "loss": 0.2787,
      "step": 3492
    },
    {
      "epoch": 2.64,
      "learning_rate": 3.349565052950076e-05,
      "loss": 0.3028,
      "step": 3493
    },
    {
      "epoch": 2.64,
      "learning_rate": 3.3490922844175493e-05,
      "loss": 0.2653,
      "step": 3494
    },
    {
      "epoch": 2.64,
      "learning_rate": 3.348619515885023e-05,
      "loss": 0.2439,
      "step": 3495
    },
    {
      "epoch": 2.64,
      "learning_rate": 3.3481467473524964e-05,
      "loss": 0.2906,
      "step": 3496
    },
    {
      "epoch": 2.64,
      "learning_rate": 3.34767397881997e-05,
      "loss": 0.3111,
      "step": 3497
    },
    {
      "epoch": 2.65,
      "learning_rate": 3.3472012102874435e-05,
      "loss": 0.2791,
      "step": 3498
    },
    {
      "epoch": 2.65,
      "learning_rate": 3.3467284417549174e-05,
      "loss": 0.3213,
      "step": 3499
    },
    {
      "epoch": 2.65,
      "learning_rate": 3.3462556732223906e-05,
      "loss": 0.2422,
      "step": 3500
    },
    {
      "epoch": 2.65,
      "learning_rate": 3.3457829046898645e-05,
      "loss": 0.3145,
      "step": 3501
    },
    {
      "epoch": 2.65,
      "learning_rate": 3.345310136157338e-05,
      "loss": 0.2804,
      "step": 3502
    },
    {
      "epoch": 2.65,
      "learning_rate": 3.3448373676248116e-05,
      "loss": 0.2789,
      "step": 3503
    },
    {
      "epoch": 2.65,
      "learning_rate": 3.344364599092285e-05,
      "loss": 0.3353,
      "step": 3504
    },
    {
      "epoch": 2.65,
      "learning_rate": 3.343891830559758e-05,
      "loss": 0.3051,
      "step": 3505
    },
    {
      "epoch": 2.65,
      "learning_rate": 3.343419062027231e-05,
      "loss": 0.2676,
      "step": 3506
    },
    {
      "epoch": 2.65,
      "learning_rate": 3.342946293494705e-05,
      "loss": 0.2563,
      "step": 3507
    },
    {
      "epoch": 2.65,
      "learning_rate": 3.342473524962178e-05,
      "loss": 0.3013,
      "step": 3508
    },
    {
      "epoch": 2.65,
      "learning_rate": 3.342000756429652e-05,
      "loss": 0.2848,
      "step": 3509
    },
    {
      "epoch": 2.65,
      "learning_rate": 3.3415279878971254e-05,
      "loss": 0.2707,
      "step": 3510
    },
    {
      "epoch": 2.66,
      "learning_rate": 3.341055219364599e-05,
      "loss": 0.2481,
      "step": 3511
    },
    {
      "epoch": 2.66,
      "learning_rate": 3.3405824508320725e-05,
      "loss": 0.3128,
      "step": 3512
    },
    {
      "epoch": 2.66,
      "learning_rate": 3.3401096822995464e-05,
      "loss": 0.2528,
      "step": 3513
    },
    {
      "epoch": 2.66,
      "learning_rate": 3.3396369137670196e-05,
      "loss": 0.2498,
      "step": 3514
    },
    {
      "epoch": 2.66,
      "learning_rate": 3.3391641452344934e-05,
      "loss": 0.2568,
      "step": 3515
    },
    {
      "epoch": 2.66,
      "learning_rate": 3.3386913767019667e-05,
      "loss": 0.2525,
      "step": 3516
    },
    {
      "epoch": 2.66,
      "learning_rate": 3.3382186081694405e-05,
      "loss": 0.3106,
      "step": 3517
    },
    {
      "epoch": 2.66,
      "learning_rate": 3.337745839636914e-05,
      "loss": 0.2697,
      "step": 3518
    },
    {
      "epoch": 2.66,
      "learning_rate": 3.3372730711043876e-05,
      "loss": 0.2906,
      "step": 3519
    },
    {
      "epoch": 2.66,
      "learning_rate": 3.336800302571861e-05,
      "loss": 0.2738,
      "step": 3520
    },
    {
      "epoch": 2.66,
      "learning_rate": 3.336327534039335e-05,
      "loss": 0.2566,
      "step": 3521
    },
    {
      "epoch": 2.66,
      "learning_rate": 3.335854765506808e-05,
      "loss": 0.2956,
      "step": 3522
    },
    {
      "epoch": 2.66,
      "learning_rate": 3.335381996974282e-05,
      "loss": 0.3084,
      "step": 3523
    },
    {
      "epoch": 2.67,
      "learning_rate": 3.334909228441755e-05,
      "loss": 0.2547,
      "step": 3524
    },
    {
      "epoch": 2.67,
      "learning_rate": 3.334436459909229e-05,
      "loss": 0.2978,
      "step": 3525
    },
    {
      "epoch": 2.67,
      "learning_rate": 3.333963691376702e-05,
      "loss": 0.3062,
      "step": 3526
    },
    {
      "epoch": 2.67,
      "learning_rate": 3.333490922844176e-05,
      "loss": 0.2896,
      "step": 3527
    },
    {
      "epoch": 2.67,
      "learning_rate": 3.333018154311649e-05,
      "loss": 0.2748,
      "step": 3528
    },
    {
      "epoch": 2.67,
      "learning_rate": 3.332545385779123e-05,
      "loss": 0.3086,
      "step": 3529
    },
    {
      "epoch": 2.67,
      "learning_rate": 3.332072617246596e-05,
      "loss": 0.3056,
      "step": 3530
    },
    {
      "epoch": 2.67,
      "learning_rate": 3.33159984871407e-05,
      "loss": 0.2711,
      "step": 3531
    },
    {
      "epoch": 2.67,
      "learning_rate": 3.3311270801815434e-05,
      "loss": 0.282,
      "step": 3532
    },
    {
      "epoch": 2.67,
      "learning_rate": 3.3306543116490166e-05,
      "loss": 0.2737,
      "step": 3533
    },
    {
      "epoch": 2.67,
      "learning_rate": 3.33018154311649e-05,
      "loss": 0.3109,
      "step": 3534
    },
    {
      "epoch": 2.67,
      "learning_rate": 3.329708774583964e-05,
      "loss": 0.2426,
      "step": 3535
    },
    {
      "epoch": 2.67,
      "learning_rate": 3.329236006051437e-05,
      "loss": 0.2806,
      "step": 3536
    },
    {
      "epoch": 2.67,
      "learning_rate": 3.328763237518911e-05,
      "loss": 0.2675,
      "step": 3537
    },
    {
      "epoch": 2.68,
      "learning_rate": 3.328290468986384e-05,
      "loss": 0.2809,
      "step": 3538
    },
    {
      "epoch": 2.68,
      "learning_rate": 3.327817700453858e-05,
      "loss": 0.2803,
      "step": 3539
    },
    {
      "epoch": 2.68,
      "learning_rate": 3.327344931921331e-05,
      "loss": 0.2905,
      "step": 3540
    },
    {
      "epoch": 2.68,
      "learning_rate": 3.326872163388805e-05,
      "loss": 0.3435,
      "step": 3541
    },
    {
      "epoch": 2.68,
      "learning_rate": 3.326399394856278e-05,
      "loss": 0.2646,
      "step": 3542
    },
    {
      "epoch": 2.68,
      "learning_rate": 3.325926626323752e-05,
      "loss": 0.286,
      "step": 3543
    },
    {
      "epoch": 2.68,
      "learning_rate": 3.325453857791225e-05,
      "loss": 0.2973,
      "step": 3544
    },
    {
      "epoch": 2.68,
      "learning_rate": 3.324981089258699e-05,
      "loss": 0.2899,
      "step": 3545
    },
    {
      "epoch": 2.68,
      "learning_rate": 3.324508320726172e-05,
      "loss": 0.2386,
      "step": 3546
    },
    {
      "epoch": 2.68,
      "learning_rate": 3.324035552193646e-05,
      "loss": 0.2757,
      "step": 3547
    },
    {
      "epoch": 2.68,
      "learning_rate": 3.32356278366112e-05,
      "loss": 0.3077,
      "step": 3548
    },
    {
      "epoch": 2.68,
      "learning_rate": 3.323090015128593e-05,
      "loss": 0.2805,
      "step": 3549
    },
    {
      "epoch": 2.68,
      "learning_rate": 3.322617246596067e-05,
      "loss": 0.2795,
      "step": 3550
    },
    {
      "epoch": 2.69,
      "learning_rate": 3.3221444780635404e-05,
      "loss": 0.2828,
      "step": 3551
    },
    {
      "epoch": 2.69,
      "learning_rate": 3.321671709531014e-05,
      "loss": 0.242,
      "step": 3552
    },
    {
      "epoch": 2.69,
      "learning_rate": 3.3211989409984875e-05,
      "loss": 0.2927,
      "step": 3553
    },
    {
      "epoch": 2.69,
      "learning_rate": 3.3207261724659614e-05,
      "loss": 0.3064,
      "step": 3554
    },
    {
      "epoch": 2.69,
      "learning_rate": 3.3202534039334346e-05,
      "loss": 0.3094,
      "step": 3555
    },
    {
      "epoch": 2.69,
      "learning_rate": 3.3197806354009084e-05,
      "loss": 0.2559,
      "step": 3556
    },
    {
      "epoch": 2.69,
      "learning_rate": 3.3193078668683816e-05,
      "loss": 0.2359,
      "step": 3557
    },
    {
      "epoch": 2.69,
      "learning_rate": 3.318835098335855e-05,
      "loss": 0.2766,
      "step": 3558
    },
    {
      "epoch": 2.69,
      "learning_rate": 3.318362329803328e-05,
      "loss": 0.2482,
      "step": 3559
    },
    {
      "epoch": 2.69,
      "learning_rate": 3.317889561270802e-05,
      "loss": 0.3022,
      "step": 3560
    },
    {
      "epoch": 2.69,
      "learning_rate": 3.317416792738275e-05,
      "loss": 0.2788,
      "step": 3561
    },
    {
      "epoch": 2.69,
      "learning_rate": 3.316944024205749e-05,
      "loss": 0.2562,
      "step": 3562
    },
    {
      "epoch": 2.69,
      "learning_rate": 3.316471255673222e-05,
      "loss": 0.2775,
      "step": 3563
    },
    {
      "epoch": 2.7,
      "learning_rate": 3.315998487140696e-05,
      "loss": 0.2997,
      "step": 3564
    },
    {
      "epoch": 2.7,
      "learning_rate": 3.315525718608169e-05,
      "loss": 0.2415,
      "step": 3565
    },
    {
      "epoch": 2.7,
      "learning_rate": 3.315052950075643e-05,
      "loss": 0.2895,
      "step": 3566
    },
    {
      "epoch": 2.7,
      "learning_rate": 3.3145801815431164e-05,
      "loss": 0.2789,
      "step": 3567
    },
    {
      "epoch": 2.7,
      "learning_rate": 3.31410741301059e-05,
      "loss": 0.2806,
      "step": 3568
    },
    {
      "epoch": 2.7,
      "learning_rate": 3.3136346444780635e-05,
      "loss": 0.3507,
      "step": 3569
    },
    {
      "epoch": 2.7,
      "learning_rate": 3.3131618759455374e-05,
      "loss": 0.3135,
      "step": 3570
    },
    {
      "epoch": 2.7,
      "learning_rate": 3.3126891074130106e-05,
      "loss": 0.2527,
      "step": 3571
    },
    {
      "epoch": 2.7,
      "learning_rate": 3.3122163388804845e-05,
      "loss": 0.2927,
      "step": 3572
    },
    {
      "epoch": 2.7,
      "learning_rate": 3.311743570347958e-05,
      "loss": 0.3153,
      "step": 3573
    },
    {
      "epoch": 2.7,
      "learning_rate": 3.3112708018154316e-05,
      "loss": 0.2733,
      "step": 3574
    },
    {
      "epoch": 2.7,
      "learning_rate": 3.310798033282905e-05,
      "loss": 0.3249,
      "step": 3575
    },
    {
      "epoch": 2.7,
      "learning_rate": 3.3103252647503787e-05,
      "loss": 0.3073,
      "step": 3576
    },
    {
      "epoch": 2.71,
      "learning_rate": 3.309852496217852e-05,
      "loss": 0.2609,
      "step": 3577
    },
    {
      "epoch": 2.71,
      "learning_rate": 3.309379727685326e-05,
      "loss": 0.2982,
      "step": 3578
    },
    {
      "epoch": 2.71,
      "learning_rate": 3.308906959152799e-05,
      "loss": 0.2861,
      "step": 3579
    },
    {
      "epoch": 2.71,
      "learning_rate": 3.308434190620273e-05,
      "loss": 0.2314,
      "step": 3580
    },
    {
      "epoch": 2.71,
      "learning_rate": 3.307961422087746e-05,
      "loss": 0.2656,
      "step": 3581
    },
    {
      "epoch": 2.71,
      "learning_rate": 3.30748865355522e-05,
      "loss": 0.2479,
      "step": 3582
    },
    {
      "epoch": 2.71,
      "learning_rate": 3.307015885022693e-05,
      "loss": 0.3222,
      "step": 3583
    },
    {
      "epoch": 2.71,
      "learning_rate": 3.306543116490167e-05,
      "loss": 0.2818,
      "step": 3584
    },
    {
      "epoch": 2.71,
      "learning_rate": 3.30607034795764e-05,
      "loss": 0.3246,
      "step": 3585
    },
    {
      "epoch": 2.71,
      "learning_rate": 3.3055975794251134e-05,
      "loss": 0.272,
      "step": 3586
    },
    {
      "epoch": 2.71,
      "learning_rate": 3.3051248108925866e-05,
      "loss": 0.2761,
      "step": 3587
    },
    {
      "epoch": 2.71,
      "learning_rate": 3.3046520423600605e-05,
      "loss": 0.2844,
      "step": 3588
    },
    {
      "epoch": 2.71,
      "learning_rate": 3.304179273827534e-05,
      "loss": 0.2806,
      "step": 3589
    },
    {
      "epoch": 2.72,
      "learning_rate": 3.3037065052950076e-05,
      "loss": 0.2541,
      "step": 3590
    },
    {
      "epoch": 2.72,
      "learning_rate": 3.303233736762481e-05,
      "loss": 0.2806,
      "step": 3591
    },
    {
      "epoch": 2.72,
      "learning_rate": 3.302760968229955e-05,
      "loss": 0.2909,
      "step": 3592
    },
    {
      "epoch": 2.72,
      "learning_rate": 3.302288199697428e-05,
      "loss": 0.3154,
      "step": 3593
    },
    {
      "epoch": 2.72,
      "learning_rate": 3.301815431164902e-05,
      "loss": 0.2689,
      "step": 3594
    },
    {
      "epoch": 2.72,
      "learning_rate": 3.301342662632375e-05,
      "loss": 0.2528,
      "step": 3595
    },
    {
      "epoch": 2.72,
      "learning_rate": 3.300869894099849e-05,
      "loss": 0.2677,
      "step": 3596
    },
    {
      "epoch": 2.72,
      "learning_rate": 3.300397125567322e-05,
      "loss": 0.3406,
      "step": 3597
    },
    {
      "epoch": 2.72,
      "learning_rate": 3.299924357034796e-05,
      "loss": 0.2492,
      "step": 3598
    },
    {
      "epoch": 2.72,
      "learning_rate": 3.299451588502269e-05,
      "loss": 0.3108,
      "step": 3599
    },
    {
      "epoch": 2.72,
      "learning_rate": 3.298978819969743e-05,
      "loss": 0.2899,
      "step": 3600
    },
    {
      "epoch": 2.72,
      "learning_rate": 3.298506051437216e-05,
      "loss": 0.2422,
      "step": 3601
    },
    {
      "epoch": 2.72,
      "learning_rate": 3.29803328290469e-05,
      "loss": 0.2949,
      "step": 3602
    },
    {
      "epoch": 2.72,
      "learning_rate": 3.2975605143721634e-05,
      "loss": 0.2755,
      "step": 3603
    },
    {
      "epoch": 2.73,
      "learning_rate": 3.297087745839637e-05,
      "loss": 0.2648,
      "step": 3604
    },
    {
      "epoch": 2.73,
      "learning_rate": 3.2966149773071104e-05,
      "loss": 0.2839,
      "step": 3605
    },
    {
      "epoch": 2.73,
      "learning_rate": 3.296142208774584e-05,
      "loss": 0.3272,
      "step": 3606
    },
    {
      "epoch": 2.73,
      "learning_rate": 3.2956694402420575e-05,
      "loss": 0.302,
      "step": 3607
    },
    {
      "epoch": 2.73,
      "learning_rate": 3.2951966717095314e-05,
      "loss": 0.2577,
      "step": 3608
    },
    {
      "epoch": 2.73,
      "learning_rate": 3.294723903177005e-05,
      "loss": 0.362,
      "step": 3609
    },
    {
      "epoch": 2.73,
      "learning_rate": 3.2942511346444785e-05,
      "loss": 0.2733,
      "step": 3610
    },
    {
      "epoch": 2.73,
      "learning_rate": 3.293778366111952e-05,
      "loss": 0.2547,
      "step": 3611
    },
    {
      "epoch": 2.73,
      "learning_rate": 3.293305597579425e-05,
      "loss": 0.3294,
      "step": 3612
    },
    {
      "epoch": 2.73,
      "learning_rate": 3.292832829046899e-05,
      "loss": 0.2698,
      "step": 3613
    },
    {
      "epoch": 2.73,
      "learning_rate": 3.292360060514372e-05,
      "loss": 0.2942,
      "step": 3614
    },
    {
      "epoch": 2.73,
      "learning_rate": 3.291887291981846e-05,
      "loss": 0.3246,
      "step": 3615
    },
    {
      "epoch": 2.73,
      "learning_rate": 3.291414523449319e-05,
      "loss": 0.2544,
      "step": 3616
    },
    {
      "epoch": 2.74,
      "learning_rate": 3.290941754916793e-05,
      "loss": 0.2515,
      "step": 3617
    },
    {
      "epoch": 2.74,
      "learning_rate": 3.290468986384266e-05,
      "loss": 0.289,
      "step": 3618
    },
    {
      "epoch": 2.74,
      "learning_rate": 3.28999621785174e-05,
      "loss": 0.2915,
      "step": 3619
    },
    {
      "epoch": 2.74,
      "learning_rate": 3.289523449319213e-05,
      "loss": 0.2702,
      "step": 3620
    },
    {
      "epoch": 2.74,
      "learning_rate": 3.289050680786687e-05,
      "loss": 0.2508,
      "step": 3621
    },
    {
      "epoch": 2.74,
      "learning_rate": 3.2885779122541604e-05,
      "loss": 0.2509,
      "step": 3622
    },
    {
      "epoch": 2.74,
      "learning_rate": 3.288105143721634e-05,
      "loss": 0.2368,
      "step": 3623
    },
    {
      "epoch": 2.74,
      "learning_rate": 3.2876323751891075e-05,
      "loss": 0.2589,
      "step": 3624
    },
    {
      "epoch": 2.74,
      "learning_rate": 3.287159606656581e-05,
      "loss": 0.2718,
      "step": 3625
    },
    {
      "epoch": 2.74,
      "learning_rate": 3.2866868381240545e-05,
      "loss": 0.306,
      "step": 3626
    },
    {
      "epoch": 2.74,
      "learning_rate": 3.2862140695915284e-05,
      "loss": 0.2819,
      "step": 3627
    },
    {
      "epoch": 2.74,
      "learning_rate": 3.2857413010590016e-05,
      "loss": 0.3178,
      "step": 3628
    },
    {
      "epoch": 2.74,
      "learning_rate": 3.2852685325264755e-05,
      "loss": 0.2672,
      "step": 3629
    },
    {
      "epoch": 2.75,
      "learning_rate": 3.284795763993949e-05,
      "loss": 0.2688,
      "step": 3630
    },
    {
      "epoch": 2.75,
      "learning_rate": 3.2843229954614226e-05,
      "loss": 0.2703,
      "step": 3631
    },
    {
      "epoch": 2.75,
      "learning_rate": 3.283850226928896e-05,
      "loss": 0.2406,
      "step": 3632
    },
    {
      "epoch": 2.75,
      "learning_rate": 3.28337745839637e-05,
      "loss": 0.2534,
      "step": 3633
    },
    {
      "epoch": 2.75,
      "learning_rate": 3.282904689863843e-05,
      "loss": 0.2632,
      "step": 3634
    },
    {
      "epoch": 2.75,
      "learning_rate": 3.282431921331317e-05,
      "loss": 0.3051,
      "step": 3635
    },
    {
      "epoch": 2.75,
      "learning_rate": 3.28195915279879e-05,
      "loss": 0.3132,
      "step": 3636
    },
    {
      "epoch": 2.75,
      "learning_rate": 3.281486384266264e-05,
      "loss": 0.2683,
      "step": 3637
    },
    {
      "epoch": 2.75,
      "learning_rate": 3.281013615733737e-05,
      "loss": 0.2788,
      "step": 3638
    },
    {
      "epoch": 2.75,
      "learning_rate": 3.28054084720121e-05,
      "loss": 0.3085,
      "step": 3639
    },
    {
      "epoch": 2.75,
      "learning_rate": 3.2800680786686835e-05,
      "loss": 0.3037,
      "step": 3640
    },
    {
      "epoch": 2.75,
      "learning_rate": 3.2795953101361574e-05,
      "loss": 0.2814,
      "step": 3641
    },
    {
      "epoch": 2.75,
      "learning_rate": 3.2791225416036306e-05,
      "loss": 0.2875,
      "step": 3642
    },
    {
      "epoch": 2.76,
      "learning_rate": 3.2786497730711045e-05,
      "loss": 0.2565,
      "step": 3643
    },
    {
      "epoch": 2.76,
      "learning_rate": 3.278177004538578e-05,
      "loss": 0.2436,
      "step": 3644
    },
    {
      "epoch": 2.76,
      "learning_rate": 3.2777042360060516e-05,
      "loss": 0.2582,
      "step": 3645
    },
    {
      "epoch": 2.76,
      "learning_rate": 3.277231467473525e-05,
      "loss": 0.2942,
      "step": 3646
    },
    {
      "epoch": 2.76,
      "learning_rate": 3.2767586989409986e-05,
      "loss": 0.2174,
      "step": 3647
    },
    {
      "epoch": 2.76,
      "learning_rate": 3.276285930408472e-05,
      "loss": 0.2345,
      "step": 3648
    },
    {
      "epoch": 2.76,
      "learning_rate": 3.275813161875946e-05,
      "loss": 0.2784,
      "step": 3649
    },
    {
      "epoch": 2.76,
      "learning_rate": 3.275340393343419e-05,
      "loss": 0.2666,
      "step": 3650
    },
    {
      "epoch": 2.76,
      "learning_rate": 3.274867624810893e-05,
      "loss": 0.2792,
      "step": 3651
    },
    {
      "epoch": 2.76,
      "learning_rate": 3.274394856278366e-05,
      "loss": 0.2971,
      "step": 3652
    },
    {
      "epoch": 2.76,
      "learning_rate": 3.27392208774584e-05,
      "loss": 0.2829,
      "step": 3653
    },
    {
      "epoch": 2.76,
      "learning_rate": 3.273449319213313e-05,
      "loss": 0.2831,
      "step": 3654
    },
    {
      "epoch": 2.76,
      "learning_rate": 3.272976550680787e-05,
      "loss": 0.2855,
      "step": 3655
    },
    {
      "epoch": 2.76,
      "learning_rate": 3.27250378214826e-05,
      "loss": 0.2731,
      "step": 3656
    },
    {
      "epoch": 2.77,
      "learning_rate": 3.272031013615734e-05,
      "loss": 0.3351,
      "step": 3657
    },
    {
      "epoch": 2.77,
      "learning_rate": 3.271558245083207e-05,
      "loss": 0.2716,
      "step": 3658
    },
    {
      "epoch": 2.77,
      "learning_rate": 3.271085476550681e-05,
      "loss": 0.2619,
      "step": 3659
    },
    {
      "epoch": 2.77,
      "learning_rate": 3.2706127080181544e-05,
      "loss": 0.355,
      "step": 3660
    },
    {
      "epoch": 2.77,
      "learning_rate": 3.270139939485628e-05,
      "loss": 0.2773,
      "step": 3661
    },
    {
      "epoch": 2.77,
      "learning_rate": 3.2696671709531015e-05,
      "loss": 0.308,
      "step": 3662
    },
    {
      "epoch": 2.77,
      "learning_rate": 3.2691944024205754e-05,
      "loss": 0.3027,
      "step": 3663
    },
    {
      "epoch": 2.77,
      "learning_rate": 3.2687216338880486e-05,
      "loss": 0.3016,
      "step": 3664
    },
    {
      "epoch": 2.77,
      "learning_rate": 3.268248865355522e-05,
      "loss": 0.2942,
      "step": 3665
    },
    {
      "epoch": 2.77,
      "learning_rate": 3.2677760968229957e-05,
      "loss": 0.2559,
      "step": 3666
    },
    {
      "epoch": 2.77,
      "learning_rate": 3.267303328290469e-05,
      "loss": 0.2802,
      "step": 3667
    },
    {
      "epoch": 2.77,
      "learning_rate": 3.266830559757943e-05,
      "loss": 0.3143,
      "step": 3668
    },
    {
      "epoch": 2.77,
      "learning_rate": 3.266357791225416e-05,
      "loss": 0.2678,
      "step": 3669
    },
    {
      "epoch": 2.78,
      "learning_rate": 3.26588502269289e-05,
      "loss": 0.28,
      "step": 3670
    },
    {
      "epoch": 2.78,
      "learning_rate": 3.265412254160363e-05,
      "loss": 0.2966,
      "step": 3671
    },
    {
      "epoch": 2.78,
      "learning_rate": 3.264939485627837e-05,
      "loss": 0.2721,
      "step": 3672
    },
    {
      "epoch": 2.78,
      "learning_rate": 3.26446671709531e-05,
      "loss": 0.2595,
      "step": 3673
    },
    {
      "epoch": 2.78,
      "learning_rate": 3.263993948562784e-05,
      "loss": 0.2692,
      "step": 3674
    },
    {
      "epoch": 2.78,
      "learning_rate": 3.263521180030257e-05,
      "loss": 0.2966,
      "step": 3675
    },
    {
      "epoch": 2.78,
      "learning_rate": 3.263048411497731e-05,
      "loss": 0.2258,
      "step": 3676
    },
    {
      "epoch": 2.78,
      "learning_rate": 3.262575642965204e-05,
      "loss": 0.2579,
      "step": 3677
    },
    {
      "epoch": 2.78,
      "learning_rate": 3.262102874432678e-05,
      "loss": 0.3029,
      "step": 3678
    },
    {
      "epoch": 2.78,
      "learning_rate": 3.2616301059001514e-05,
      "loss": 0.2777,
      "step": 3679
    },
    {
      "epoch": 2.78,
      "learning_rate": 3.261157337367625e-05,
      "loss": 0.2953,
      "step": 3680
    },
    {
      "epoch": 2.78,
      "learning_rate": 3.2606845688350985e-05,
      "loss": 0.288,
      "step": 3681
    },
    {
      "epoch": 2.78,
      "learning_rate": 3.2602118003025724e-05,
      "loss": 0.3052,
      "step": 3682
    },
    {
      "epoch": 2.79,
      "learning_rate": 3.2597390317700456e-05,
      "loss": 0.3111,
      "step": 3683
    },
    {
      "epoch": 2.79,
      "learning_rate": 3.2592662632375195e-05,
      "loss": 0.2743,
      "step": 3684
    },
    {
      "epoch": 2.79,
      "learning_rate": 3.258793494704993e-05,
      "loss": 0.3096,
      "step": 3685
    },
    {
      "epoch": 2.79,
      "learning_rate": 3.2583207261724665e-05,
      "loss": 0.2327,
      "step": 3686
    },
    {
      "epoch": 2.79,
      "learning_rate": 3.25784795763994e-05,
      "loss": 0.2383,
      "step": 3687
    },
    {
      "epoch": 2.79,
      "learning_rate": 3.2573751891074136e-05,
      "loss": 0.2971,
      "step": 3688
    },
    {
      "epoch": 2.79,
      "learning_rate": 3.256902420574887e-05,
      "loss": 0.2907,
      "step": 3689
    },
    {
      "epoch": 2.79,
      "learning_rate": 3.256429652042361e-05,
      "loss": 0.2597,
      "step": 3690
    },
    {
      "epoch": 2.79,
      "learning_rate": 3.255956883509834e-05,
      "loss": 0.2874,
      "step": 3691
    },
    {
      "epoch": 2.79,
      "learning_rate": 3.255484114977307e-05,
      "loss": 0.3437,
      "step": 3692
    },
    {
      "epoch": 2.79,
      "learning_rate": 3.2550113464447803e-05,
      "loss": 0.3165,
      "step": 3693
    },
    {
      "epoch": 2.79,
      "learning_rate": 3.254538577912254e-05,
      "loss": 0.2551,
      "step": 3694
    },
    {
      "epoch": 2.79,
      "learning_rate": 3.2540658093797274e-05,
      "loss": 0.2751,
      "step": 3695
    },
    {
      "epoch": 2.8,
      "learning_rate": 3.253593040847201e-05,
      "loss": 0.264,
      "step": 3696
    },
    {
      "epoch": 2.8,
      "learning_rate": 3.2531202723146745e-05,
      "loss": 0.2439,
      "step": 3697
    },
    {
      "epoch": 2.8,
      "learning_rate": 3.2526475037821484e-05,
      "loss": 0.2653,
      "step": 3698
    },
    {
      "epoch": 2.8,
      "learning_rate": 3.2521747352496216e-05,
      "loss": 0.3125,
      "step": 3699
    },
    {
      "epoch": 2.8,
      "learning_rate": 3.2517019667170955e-05,
      "loss": 0.3439,
      "step": 3700
    },
    {
      "epoch": 2.8,
      "learning_rate": 3.251229198184569e-05,
      "loss": 0.2994,
      "step": 3701
    },
    {
      "epoch": 2.8,
      "learning_rate": 3.2507564296520426e-05,
      "loss": 0.2823,
      "step": 3702
    },
    {
      "epoch": 2.8,
      "learning_rate": 3.250283661119516e-05,
      "loss": 0.2682,
      "step": 3703
    },
    {
      "epoch": 2.8,
      "learning_rate": 3.24981089258699e-05,
      "loss": 0.2762,
      "step": 3704
    },
    {
      "epoch": 2.8,
      "learning_rate": 3.249338124054463e-05,
      "loss": 0.276,
      "step": 3705
    },
    {
      "epoch": 2.8,
      "learning_rate": 3.248865355521937e-05,
      "loss": 0.2689,
      "step": 3706
    },
    {
      "epoch": 2.8,
      "learning_rate": 3.24839258698941e-05,
      "loss": 0.2807,
      "step": 3707
    },
    {
      "epoch": 2.8,
      "learning_rate": 3.247919818456884e-05,
      "loss": 0.2699,
      "step": 3708
    },
    {
      "epoch": 2.81,
      "learning_rate": 3.247447049924357e-05,
      "loss": 0.2795,
      "step": 3709
    },
    {
      "epoch": 2.81,
      "learning_rate": 3.246974281391831e-05,
      "loss": 0.3053,
      "step": 3710
    },
    {
      "epoch": 2.81,
      "learning_rate": 3.246501512859304e-05,
      "loss": 0.2575,
      "step": 3711
    },
    {
      "epoch": 2.81,
      "learning_rate": 3.246028744326778e-05,
      "loss": 0.2904,
      "step": 3712
    },
    {
      "epoch": 2.81,
      "learning_rate": 3.245555975794251e-05,
      "loss": 0.2408,
      "step": 3713
    },
    {
      "epoch": 2.81,
      "learning_rate": 3.245083207261725e-05,
      "loss": 0.2546,
      "step": 3714
    },
    {
      "epoch": 2.81,
      "learning_rate": 3.244610438729198e-05,
      "loss": 0.3368,
      "step": 3715
    },
    {
      "epoch": 2.81,
      "learning_rate": 3.244137670196672e-05,
      "loss": 0.3014,
      "step": 3716
    },
    {
      "epoch": 2.81,
      "learning_rate": 3.2436649016641454e-05,
      "loss": 0.2913,
      "step": 3717
    },
    {
      "epoch": 2.81,
      "learning_rate": 3.2431921331316186e-05,
      "loss": 0.2663,
      "step": 3718
    },
    {
      "epoch": 2.81,
      "learning_rate": 3.2427193645990925e-05,
      "loss": 0.2785,
      "step": 3719
    },
    {
      "epoch": 2.81,
      "learning_rate": 3.242246596066566e-05,
      "loss": 0.2751,
      "step": 3720
    },
    {
      "epoch": 2.81,
      "learning_rate": 3.241773827534039e-05,
      "loss": 0.3026,
      "step": 3721
    },
    {
      "epoch": 2.81,
      "learning_rate": 3.241301059001513e-05,
      "loss": 0.2843,
      "step": 3722
    },
    {
      "epoch": 2.82,
      "learning_rate": 3.240828290468986e-05,
      "loss": 0.241,
      "step": 3723
    },
    {
      "epoch": 2.82,
      "learning_rate": 3.24035552193646e-05,
      "loss": 0.3097,
      "step": 3724
    },
    {
      "epoch": 2.82,
      "learning_rate": 3.239882753403933e-05,
      "loss": 0.2872,
      "step": 3725
    },
    {
      "epoch": 2.82,
      "learning_rate": 3.239409984871407e-05,
      "loss": 0.3163,
      "step": 3726
    },
    {
      "epoch": 2.82,
      "learning_rate": 3.23893721633888e-05,
      "loss": 0.2614,
      "step": 3727
    },
    {
      "epoch": 2.82,
      "learning_rate": 3.238464447806354e-05,
      "loss": 0.2641,
      "step": 3728
    },
    {
      "epoch": 2.82,
      "learning_rate": 3.237991679273828e-05,
      "loss": 0.2686,
      "step": 3729
    },
    {
      "epoch": 2.82,
      "learning_rate": 3.237518910741301e-05,
      "loss": 0.2643,
      "step": 3730
    },
    {
      "epoch": 2.82,
      "learning_rate": 3.237046142208775e-05,
      "loss": 0.2702,
      "step": 3731
    },
    {
      "epoch": 2.82,
      "learning_rate": 3.236573373676248e-05,
      "loss": 0.2895,
      "step": 3732
    },
    {
      "epoch": 2.82,
      "learning_rate": 3.236100605143722e-05,
      "loss": 0.3041,
      "step": 3733
    },
    {
      "epoch": 2.82,
      "learning_rate": 3.2356278366111953e-05,
      "loss": 0.3267,
      "step": 3734
    },
    {
      "epoch": 2.82,
      "learning_rate": 3.235155068078669e-05,
      "loss": 0.2499,
      "step": 3735
    },
    {
      "epoch": 2.83,
      "learning_rate": 3.2346822995461424e-05,
      "loss": 0.2766,
      "step": 3736
    },
    {
      "epoch": 2.83,
      "learning_rate": 3.234209531013616e-05,
      "loss": 0.2698,
      "step": 3737
    },
    {
      "epoch": 2.83,
      "learning_rate": 3.2337367624810895e-05,
      "loss": 0.345,
      "step": 3738
    },
    {
      "epoch": 2.83,
      "learning_rate": 3.2332639939485634e-05,
      "loss": 0.2824,
      "step": 3739
    },
    {
      "epoch": 2.83,
      "learning_rate": 3.2327912254160366e-05,
      "loss": 0.2737,
      "step": 3740
    },
    {
      "epoch": 2.83,
      "learning_rate": 3.2323184568835105e-05,
      "loss": 0.2909,
      "step": 3741
    },
    {
      "epoch": 2.83,
      "learning_rate": 3.231845688350984e-05,
      "loss": 0.2524,
      "step": 3742
    },
    {
      "epoch": 2.83,
      "learning_rate": 3.2313729198184576e-05,
      "loss": 0.2907,
      "step": 3743
    },
    {
      "epoch": 2.83,
      "learning_rate": 3.230900151285931e-05,
      "loss": 0.301,
      "step": 3744
    },
    {
      "epoch": 2.83,
      "learning_rate": 3.230427382753404e-05,
      "loss": 0.2487,
      "step": 3745
    },
    {
      "epoch": 2.83,
      "learning_rate": 3.229954614220877e-05,
      "loss": 0.2768,
      "step": 3746
    },
    {
      "epoch": 2.83,
      "learning_rate": 3.229481845688351e-05,
      "loss": 0.2639,
      "step": 3747
    },
    {
      "epoch": 2.83,
      "learning_rate": 3.229009077155824e-05,
      "loss": 0.308,
      "step": 3748
    },
    {
      "epoch": 2.84,
      "learning_rate": 3.228536308623298e-05,
      "loss": 0.2502,
      "step": 3749
    },
    {
      "epoch": 2.84,
      "learning_rate": 3.2280635400907714e-05,
      "loss": 0.3054,
      "step": 3750
    },
    {
      "epoch": 2.84,
      "learning_rate": 3.227590771558245e-05,
      "loss": 0.2442,
      "step": 3751
    },
    {
      "epoch": 2.84,
      "learning_rate": 3.2271180030257185e-05,
      "loss": 0.2999,
      "step": 3752
    },
    {
      "epoch": 2.84,
      "learning_rate": 3.2266452344931924e-05,
      "loss": 0.2658,
      "step": 3753
    },
    {
      "epoch": 2.84,
      "learning_rate": 3.2261724659606656e-05,
      "loss": 0.3336,
      "step": 3754
    },
    {
      "epoch": 2.84,
      "learning_rate": 3.2256996974281394e-05,
      "loss": 0.2831,
      "step": 3755
    },
    {
      "epoch": 2.84,
      "learning_rate": 3.2252269288956126e-05,
      "loss": 0.2616,
      "step": 3756
    },
    {
      "epoch": 2.84,
      "learning_rate": 3.2247541603630865e-05,
      "loss": 0.2921,
      "step": 3757
    },
    {
      "epoch": 2.84,
      "learning_rate": 3.22428139183056e-05,
      "loss": 0.261,
      "step": 3758
    },
    {
      "epoch": 2.84,
      "learning_rate": 3.2238086232980336e-05,
      "loss": 0.2606,
      "step": 3759
    },
    {
      "epoch": 2.84,
      "learning_rate": 3.223335854765507e-05,
      "loss": 0.2759,
      "step": 3760
    },
    {
      "epoch": 2.84,
      "learning_rate": 3.222863086232981e-05,
      "loss": 0.2651,
      "step": 3761
    },
    {
      "epoch": 2.85,
      "learning_rate": 3.222390317700454e-05,
      "loss": 0.2946,
      "step": 3762
    },
    {
      "epoch": 2.85,
      "learning_rate": 3.221917549167928e-05,
      "loss": 0.3133,
      "step": 3763
    },
    {
      "epoch": 2.85,
      "learning_rate": 3.221444780635401e-05,
      "loss": 0.2983,
      "step": 3764
    },
    {
      "epoch": 2.85,
      "learning_rate": 3.220972012102875e-05,
      "loss": 0.2358,
      "step": 3765
    },
    {
      "epoch": 2.85,
      "learning_rate": 3.220499243570348e-05,
      "loss": 0.2576,
      "step": 3766
    },
    {
      "epoch": 2.85,
      "learning_rate": 3.220026475037822e-05,
      "loss": 0.3063,
      "step": 3767
    },
    {
      "epoch": 2.85,
      "learning_rate": 3.219553706505295e-05,
      "loss": 0.3093,
      "step": 3768
    },
    {
      "epoch": 2.85,
      "learning_rate": 3.219080937972769e-05,
      "loss": 0.3209,
      "step": 3769
    },
    {
      "epoch": 2.85,
      "learning_rate": 3.218608169440242e-05,
      "loss": 0.2617,
      "step": 3770
    },
    {
      "epoch": 2.85,
      "learning_rate": 3.2181354009077155e-05,
      "loss": 0.285,
      "step": 3771
    },
    {
      "epoch": 2.85,
      "learning_rate": 3.2176626323751894e-05,
      "loss": 0.2711,
      "step": 3772
    },
    {
      "epoch": 2.85,
      "learning_rate": 3.2171898638426626e-05,
      "loss": 0.2571,
      "step": 3773
    },
    {
      "epoch": 2.85,
      "learning_rate": 3.216717095310136e-05,
      "loss": 0.3146,
      "step": 3774
    },
    {
      "epoch": 2.85,
      "learning_rate": 3.2162443267776097e-05,
      "loss": 0.2949,
      "step": 3775
    },
    {
      "epoch": 2.86,
      "learning_rate": 3.215771558245083e-05,
      "loss": 0.3022,
      "step": 3776
    },
    {
      "epoch": 2.86,
      "learning_rate": 3.215298789712557e-05,
      "loss": 0.2961,
      "step": 3777
    },
    {
      "epoch": 2.86,
      "learning_rate": 3.21482602118003e-05,
      "loss": 0.2853,
      "step": 3778
    },
    {
      "epoch": 2.86,
      "learning_rate": 3.214353252647504e-05,
      "loss": 0.2563,
      "step": 3779
    },
    {
      "epoch": 2.86,
      "learning_rate": 3.213880484114977e-05,
      "loss": 0.2685,
      "step": 3780
    },
    {
      "epoch": 2.86,
      "learning_rate": 3.213407715582451e-05,
      "loss": 0.2903,
      "step": 3781
    },
    {
      "epoch": 2.86,
      "learning_rate": 3.212934947049924e-05,
      "loss": 0.2942,
      "step": 3782
    },
    {
      "epoch": 2.86,
      "learning_rate": 3.212462178517398e-05,
      "loss": 0.2951,
      "step": 3783
    },
    {
      "epoch": 2.86,
      "learning_rate": 3.211989409984871e-05,
      "loss": 0.3334,
      "step": 3784
    },
    {
      "epoch": 2.86,
      "learning_rate": 3.211516641452345e-05,
      "loss": 0.2916,
      "step": 3785
    },
    {
      "epoch": 2.86,
      "learning_rate": 3.211043872919818e-05,
      "loss": 0.286,
      "step": 3786
    },
    {
      "epoch": 2.86,
      "learning_rate": 3.210571104387292e-05,
      "loss": 0.249,
      "step": 3787
    },
    {
      "epoch": 2.86,
      "learning_rate": 3.210098335854766e-05,
      "loss": 0.2981,
      "step": 3788
    },
    {
      "epoch": 2.87,
      "learning_rate": 3.209625567322239e-05,
      "loss": 0.268,
      "step": 3789
    },
    {
      "epoch": 2.87,
      "learning_rate": 3.209152798789713e-05,
      "loss": 0.3108,
      "step": 3790
    },
    {
      "epoch": 2.87,
      "learning_rate": 3.2086800302571864e-05,
      "loss": 0.3138,
      "step": 3791
    },
    {
      "epoch": 2.87,
      "learning_rate": 3.20820726172466e-05,
      "loss": 0.2681,
      "step": 3792
    },
    {
      "epoch": 2.87,
      "learning_rate": 3.2077344931921335e-05,
      "loss": 0.2663,
      "step": 3793
    },
    {
      "epoch": 2.87,
      "learning_rate": 3.2072617246596073e-05,
      "loss": 0.2669,
      "step": 3794
    },
    {
      "epoch": 2.87,
      "learning_rate": 3.2067889561270806e-05,
      "loss": 0.2403,
      "step": 3795
    },
    {
      "epoch": 2.87,
      "learning_rate": 3.2063161875945544e-05,
      "loss": 0.2849,
      "step": 3796
    },
    {
      "epoch": 2.87,
      "learning_rate": 3.2058434190620276e-05,
      "loss": 0.2785,
      "step": 3797
    },
    {
      "epoch": 2.87,
      "learning_rate": 3.205370650529501e-05,
      "loss": 0.254,
      "step": 3798
    },
    {
      "epoch": 2.87,
      "learning_rate": 3.204897881996974e-05,
      "loss": 0.2611,
      "step": 3799
    },
    {
      "epoch": 2.87,
      "learning_rate": 3.204425113464448e-05,
      "loss": 0.2609,
      "step": 3800
    },
    {
      "epoch": 2.87,
      "learning_rate": 3.203952344931921e-05,
      "loss": 0.2642,
      "step": 3801
    },
    {
      "epoch": 2.88,
      "learning_rate": 3.203479576399395e-05,
      "loss": 0.2987,
      "step": 3802
    },
    {
      "epoch": 2.88,
      "learning_rate": 3.203006807866868e-05,
      "loss": 0.2898,
      "step": 3803
    },
    {
      "epoch": 2.88,
      "learning_rate": 3.202534039334342e-05,
      "loss": 0.2928,
      "step": 3804
    },
    {
      "epoch": 2.88,
      "learning_rate": 3.202061270801815e-05,
      "loss": 0.3077,
      "step": 3805
    },
    {
      "epoch": 2.88,
      "learning_rate": 3.201588502269289e-05,
      "loss": 0.2754,
      "step": 3806
    },
    {
      "epoch": 2.88,
      "learning_rate": 3.2011157337367624e-05,
      "loss": 0.3463,
      "step": 3807
    },
    {
      "epoch": 2.88,
      "learning_rate": 3.200642965204236e-05,
      "loss": 0.2883,
      "step": 3808
    },
    {
      "epoch": 2.88,
      "learning_rate": 3.2001701966717095e-05,
      "loss": 0.2537,
      "step": 3809
    },
    {
      "epoch": 2.88,
      "learning_rate": 3.1996974281391834e-05,
      "loss": 0.2904,
      "step": 3810
    },
    {
      "epoch": 2.88,
      "learning_rate": 3.1992246596066566e-05,
      "loss": 0.2793,
      "step": 3811
    },
    {
      "epoch": 2.88,
      "learning_rate": 3.1987518910741305e-05,
      "loss": 0.2959,
      "step": 3812
    },
    {
      "epoch": 2.88,
      "learning_rate": 3.198279122541604e-05,
      "loss": 0.2902,
      "step": 3813
    },
    {
      "epoch": 2.88,
      "learning_rate": 3.1978063540090776e-05,
      "loss": 0.2944,
      "step": 3814
    },
    {
      "epoch": 2.89,
      "learning_rate": 3.197333585476551e-05,
      "loss": 0.3028,
      "step": 3815
    },
    {
      "epoch": 2.89,
      "learning_rate": 3.1968608169440247e-05,
      "loss": 0.301,
      "step": 3816
    },
    {
      "epoch": 2.89,
      "learning_rate": 3.196388048411498e-05,
      "loss": 0.3193,
      "step": 3817
    },
    {
      "epoch": 2.89,
      "learning_rate": 3.195915279878972e-05,
      "loss": 0.2852,
      "step": 3818
    },
    {
      "epoch": 2.89,
      "learning_rate": 3.195442511346445e-05,
      "loss": 0.259,
      "step": 3819
    },
    {
      "epoch": 2.89,
      "learning_rate": 3.194969742813919e-05,
      "loss": 0.2627,
      "step": 3820
    },
    {
      "epoch": 2.89,
      "learning_rate": 3.194496974281392e-05,
      "loss": 0.2559,
      "step": 3821
    },
    {
      "epoch": 2.89,
      "learning_rate": 3.194024205748866e-05,
      "loss": 0.3083,
      "step": 3822
    },
    {
      "epoch": 2.89,
      "learning_rate": 3.193551437216339e-05,
      "loss": 0.2617,
      "step": 3823
    },
    {
      "epoch": 2.89,
      "learning_rate": 3.193078668683812e-05,
      "loss": 0.242,
      "step": 3824
    },
    {
      "epoch": 2.89,
      "learning_rate": 3.192605900151286e-05,
      "loss": 0.2899,
      "step": 3825
    },
    {
      "epoch": 2.89,
      "learning_rate": 3.1921331316187594e-05,
      "loss": 0.3157,
      "step": 3826
    },
    {
      "epoch": 2.89,
      "learning_rate": 3.1916603630862326e-05,
      "loss": 0.2334,
      "step": 3827
    },
    {
      "epoch": 2.9,
      "learning_rate": 3.1911875945537065e-05,
      "loss": 0.2851,
      "step": 3828
    },
    {
      "epoch": 2.9,
      "learning_rate": 3.19071482602118e-05,
      "loss": 0.2446,
      "step": 3829
    },
    {
      "epoch": 2.9,
      "learning_rate": 3.1902420574886536e-05,
      "loss": 0.317,
      "step": 3830
    },
    {
      "epoch": 2.9,
      "learning_rate": 3.189769288956127e-05,
      "loss": 0.2915,
      "step": 3831
    },
    {
      "epoch": 2.9,
      "learning_rate": 3.189296520423601e-05,
      "loss": 0.2845,
      "step": 3832
    },
    {
      "epoch": 2.9,
      "learning_rate": 3.188823751891074e-05,
      "loss": 0.2603,
      "step": 3833
    },
    {
      "epoch": 2.9,
      "learning_rate": 3.188350983358548e-05,
      "loss": 0.293,
      "step": 3834
    },
    {
      "epoch": 2.9,
      "learning_rate": 3.187878214826021e-05,
      "loss": 0.2962,
      "step": 3835
    },
    {
      "epoch": 2.9,
      "learning_rate": 3.187405446293495e-05,
      "loss": 0.274,
      "step": 3836
    },
    {
      "epoch": 2.9,
      "learning_rate": 3.186932677760968e-05,
      "loss": 0.3088,
      "step": 3837
    },
    {
      "epoch": 2.9,
      "learning_rate": 3.186459909228442e-05,
      "loss": 0.3055,
      "step": 3838
    },
    {
      "epoch": 2.9,
      "learning_rate": 3.185987140695915e-05,
      "loss": 0.2813,
      "step": 3839
    },
    {
      "epoch": 2.9,
      "learning_rate": 3.185514372163389e-05,
      "loss": 0.2852,
      "step": 3840
    },
    {
      "epoch": 2.9,
      "learning_rate": 3.185041603630862e-05,
      "loss": 0.2654,
      "step": 3841
    },
    {
      "epoch": 2.91,
      "learning_rate": 3.184568835098336e-05,
      "loss": 0.3376,
      "step": 3842
    },
    {
      "epoch": 2.91,
      "learning_rate": 3.1840960665658093e-05,
      "loss": 0.2863,
      "step": 3843
    },
    {
      "epoch": 2.91,
      "learning_rate": 3.183623298033283e-05,
      "loss": 0.2575,
      "step": 3844
    },
    {
      "epoch": 2.91,
      "learning_rate": 3.1831505295007564e-05,
      "loss": 0.2725,
      "step": 3845
    },
    {
      "epoch": 2.91,
      "learning_rate": 3.18267776096823e-05,
      "loss": 0.2508,
      "step": 3846
    },
    {
      "epoch": 2.91,
      "learning_rate": 3.1822049924357035e-05,
      "loss": 0.2858,
      "step": 3847
    },
    {
      "epoch": 2.91,
      "learning_rate": 3.1817322239031774e-05,
      "loss": 0.3387,
      "step": 3848
    },
    {
      "epoch": 2.91,
      "learning_rate": 3.181259455370651e-05,
      "loss": 0.285,
      "step": 3849
    },
    {
      "epoch": 2.91,
      "learning_rate": 3.1807866868381245e-05,
      "loss": 0.2854,
      "step": 3850
    },
    {
      "epoch": 2.91,
      "learning_rate": 3.180313918305598e-05,
      "loss": 0.2866,
      "step": 3851
    },
    {
      "epoch": 2.91,
      "learning_rate": 3.179841149773071e-05,
      "loss": 0.2791,
      "step": 3852
    },
    {
      "epoch": 2.91,
      "learning_rate": 3.179368381240545e-05,
      "loss": 0.2901,
      "step": 3853
    },
    {
      "epoch": 2.91,
      "learning_rate": 3.178895612708018e-05,
      "loss": 0.2848,
      "step": 3854
    },
    {
      "epoch": 2.92,
      "learning_rate": 3.178422844175492e-05,
      "loss": 0.3223,
      "step": 3855
    },
    {
      "epoch": 2.92,
      "learning_rate": 3.177950075642965e-05,
      "loss": 0.3193,
      "step": 3856
    },
    {
      "epoch": 2.92,
      "learning_rate": 3.177477307110439e-05,
      "loss": 0.2607,
      "step": 3857
    },
    {
      "epoch": 2.92,
      "learning_rate": 3.177004538577912e-05,
      "loss": 0.2647,
      "step": 3858
    },
    {
      "epoch": 2.92,
      "learning_rate": 3.176531770045386e-05,
      "loss": 0.2753,
      "step": 3859
    },
    {
      "epoch": 2.92,
      "learning_rate": 3.176059001512859e-05,
      "loss": 0.2926,
      "step": 3860
    },
    {
      "epoch": 2.92,
      "learning_rate": 3.175586232980333e-05,
      "loss": 0.2905,
      "step": 3861
    },
    {
      "epoch": 2.92,
      "learning_rate": 3.1751134644478064e-05,
      "loss": 0.2779,
      "step": 3862
    },
    {
      "epoch": 2.92,
      "learning_rate": 3.17464069591528e-05,
      "loss": 0.2576,
      "step": 3863
    },
    {
      "epoch": 2.92,
      "learning_rate": 3.1741679273827534e-05,
      "loss": 0.2933,
      "step": 3864
    },
    {
      "epoch": 2.92,
      "learning_rate": 3.173695158850227e-05,
      "loss": 0.324,
      "step": 3865
    },
    {
      "epoch": 2.92,
      "learning_rate": 3.1732223903177005e-05,
      "loss": 0.3019,
      "step": 3866
    },
    {
      "epoch": 2.92,
      "learning_rate": 3.1727496217851744e-05,
      "loss": 0.3011,
      "step": 3867
    },
    {
      "epoch": 2.93,
      "learning_rate": 3.1722768532526476e-05,
      "loss": 0.2816,
      "step": 3868
    },
    {
      "epoch": 2.93,
      "learning_rate": 3.1718040847201215e-05,
      "loss": 0.3229,
      "step": 3869
    },
    {
      "epoch": 2.93,
      "learning_rate": 3.171331316187595e-05,
      "loss": 0.2683,
      "step": 3870
    },
    {
      "epoch": 2.93,
      "learning_rate": 3.1708585476550686e-05,
      "loss": 0.2568,
      "step": 3871
    },
    {
      "epoch": 2.93,
      "learning_rate": 3.170385779122542e-05,
      "loss": 0.3055,
      "step": 3872
    },
    {
      "epoch": 2.93,
      "learning_rate": 3.169913010590016e-05,
      "loss": 0.2359,
      "step": 3873
    },
    {
      "epoch": 2.93,
      "learning_rate": 3.169440242057489e-05,
      "loss": 0.3171,
      "step": 3874
    },
    {
      "epoch": 2.93,
      "learning_rate": 3.168967473524963e-05,
      "loss": 0.2705,
      "step": 3875
    },
    {
      "epoch": 2.93,
      "learning_rate": 3.168494704992436e-05,
      "loss": 0.3116,
      "step": 3876
    },
    {
      "epoch": 2.93,
      "learning_rate": 3.168021936459909e-05,
      "loss": 0.309,
      "step": 3877
    },
    {
      "epoch": 2.93,
      "learning_rate": 3.167549167927383e-05,
      "loss": 0.2858,
      "step": 3878
    },
    {
      "epoch": 2.93,
      "learning_rate": 3.167076399394856e-05,
      "loss": 0.279,
      "step": 3879
    },
    {
      "epoch": 2.93,
      "learning_rate": 3.1666036308623295e-05,
      "loss": 0.2878,
      "step": 3880
    },
    {
      "epoch": 2.94,
      "learning_rate": 3.1661308623298034e-05,
      "loss": 0.331,
      "step": 3881
    },
    {
      "epoch": 2.94,
      "learning_rate": 3.1656580937972766e-05,
      "loss": 0.271,
      "step": 3882
    },
    {
      "epoch": 2.94,
      "learning_rate": 3.1651853252647505e-05,
      "loss": 0.2794,
      "step": 3883
    },
    {
      "epoch": 2.94,
      "learning_rate": 3.164712556732224e-05,
      "loss": 0.2743,
      "step": 3884
    },
    {
      "epoch": 2.94,
      "learning_rate": 3.1642397881996975e-05,
      "loss": 0.2768,
      "step": 3885
    },
    {
      "epoch": 2.94,
      "learning_rate": 3.163767019667171e-05,
      "loss": 0.3396,
      "step": 3886
    },
    {
      "epoch": 2.94,
      "learning_rate": 3.1632942511346446e-05,
      "loss": 0.2703,
      "step": 3887
    },
    {
      "epoch": 2.94,
      "learning_rate": 3.162821482602118e-05,
      "loss": 0.2621,
      "step": 3888
    },
    {
      "epoch": 2.94,
      "learning_rate": 3.162348714069592e-05,
      "loss": 0.2749,
      "step": 3889
    },
    {
      "epoch": 2.94,
      "learning_rate": 3.161875945537065e-05,
      "loss": 0.2664,
      "step": 3890
    },
    {
      "epoch": 2.94,
      "learning_rate": 3.161403177004539e-05,
      "loss": 0.2693,
      "step": 3891
    },
    {
      "epoch": 2.94,
      "learning_rate": 3.160930408472012e-05,
      "loss": 0.3094,
      "step": 3892
    },
    {
      "epoch": 2.94,
      "learning_rate": 3.160457639939486e-05,
      "loss": 0.2628,
      "step": 3893
    },
    {
      "epoch": 2.94,
      "learning_rate": 3.159984871406959e-05,
      "loss": 0.2859,
      "step": 3894
    },
    {
      "epoch": 2.95,
      "learning_rate": 3.159512102874433e-05,
      "loss": 0.2498,
      "step": 3895
    },
    {
      "epoch": 2.95,
      "learning_rate": 3.159039334341906e-05,
      "loss": 0.2648,
      "step": 3896
    },
    {
      "epoch": 2.95,
      "learning_rate": 3.15856656580938e-05,
      "loss": 0.2626,
      "step": 3897
    },
    {
      "epoch": 2.95,
      "learning_rate": 3.158093797276853e-05,
      "loss": 0.2626,
      "step": 3898
    },
    {
      "epoch": 2.95,
      "learning_rate": 3.157621028744327e-05,
      "loss": 0.2823,
      "step": 3899
    },
    {
      "epoch": 2.95,
      "learning_rate": 3.1571482602118004e-05,
      "loss": 0.2875,
      "step": 3900
    },
    {
      "epoch": 2.95,
      "learning_rate": 3.156675491679274e-05,
      "loss": 0.2645,
      "step": 3901
    },
    {
      "epoch": 2.95,
      "learning_rate": 3.1562027231467475e-05,
      "loss": 0.2387,
      "step": 3902
    },
    {
      "epoch": 2.95,
      "learning_rate": 3.1557299546142214e-05,
      "loss": 0.2871,
      "step": 3903
    },
    {
      "epoch": 2.95,
      "learning_rate": 3.1552571860816946e-05,
      "loss": 0.2629,
      "step": 3904
    },
    {
      "epoch": 2.95,
      "learning_rate": 3.154784417549168e-05,
      "loss": 0.2904,
      "step": 3905
    },
    {
      "epoch": 2.95,
      "learning_rate": 3.1543116490166416e-05,
      "loss": 0.2927,
      "step": 3906
    },
    {
      "epoch": 2.95,
      "learning_rate": 3.153838880484115e-05,
      "loss": 0.2911,
      "step": 3907
    },
    {
      "epoch": 2.96,
      "learning_rate": 3.153366111951589e-05,
      "loss": 0.2603,
      "step": 3908
    },
    {
      "epoch": 2.96,
      "learning_rate": 3.152893343419062e-05,
      "loss": 0.3057,
      "step": 3909
    },
    {
      "epoch": 2.96,
      "learning_rate": 3.152420574886536e-05,
      "loss": 0.2596,
      "step": 3910
    },
    {
      "epoch": 2.96,
      "learning_rate": 3.151947806354009e-05,
      "loss": 0.3245,
      "step": 3911
    },
    {
      "epoch": 2.96,
      "learning_rate": 3.151475037821483e-05,
      "loss": 0.331,
      "step": 3912
    },
    {
      "epoch": 2.96,
      "learning_rate": 3.151002269288956e-05,
      "loss": 0.2783,
      "step": 3913
    },
    {
      "epoch": 2.96,
      "learning_rate": 3.15052950075643e-05,
      "loss": 0.288,
      "step": 3914
    },
    {
      "epoch": 2.96,
      "learning_rate": 3.150056732223903e-05,
      "loss": 0.2839,
      "step": 3915
    },
    {
      "epoch": 2.96,
      "learning_rate": 3.149583963691377e-05,
      "loss": 0.2516,
      "step": 3916
    },
    {
      "epoch": 2.96,
      "learning_rate": 3.14911119515885e-05,
      "loss": 0.2451,
      "step": 3917
    },
    {
      "epoch": 2.96,
      "learning_rate": 3.148638426626324e-05,
      "loss": 0.2436,
      "step": 3918
    },
    {
      "epoch": 2.96,
      "learning_rate": 3.1481656580937974e-05,
      "loss": 0.2616,
      "step": 3919
    },
    {
      "epoch": 2.96,
      "learning_rate": 3.147692889561271e-05,
      "loss": 0.283,
      "step": 3920
    },
    {
      "epoch": 2.97,
      "learning_rate": 3.1472201210287445e-05,
      "loss": 0.3107,
      "step": 3921
    },
    {
      "epoch": 2.97,
      "learning_rate": 3.1467473524962184e-05,
      "loss": 0.2777,
      "step": 3922
    },
    {
      "epoch": 2.97,
      "learning_rate": 3.1462745839636916e-05,
      "loss": 0.2874,
      "step": 3923
    },
    {
      "epoch": 2.97,
      "learning_rate": 3.1458018154311655e-05,
      "loss": 0.2929,
      "step": 3924
    },
    {
      "epoch": 2.97,
      "learning_rate": 3.1453290468986387e-05,
      "loss": 0.2687,
      "step": 3925
    },
    {
      "epoch": 2.97,
      "learning_rate": 3.1448562783661125e-05,
      "loss": 0.2761,
      "step": 3926
    },
    {
      "epoch": 2.97,
      "learning_rate": 3.144383509833586e-05,
      "loss": 0.2549,
      "step": 3927
    },
    {
      "epoch": 2.97,
      "learning_rate": 3.1439107413010596e-05,
      "loss": 0.2485,
      "step": 3928
    },
    {
      "epoch": 2.97,
      "learning_rate": 3.143437972768533e-05,
      "loss": 0.3061,
      "step": 3929
    },
    {
      "epoch": 2.97,
      "learning_rate": 3.142965204236006e-05,
      "loss": 0.2946,
      "step": 3930
    },
    {
      "epoch": 2.97,
      "learning_rate": 3.14249243570348e-05,
      "loss": 0.275,
      "step": 3931
    },
    {
      "epoch": 2.97,
      "learning_rate": 3.142019667170953e-05,
      "loss": 0.2895,
      "step": 3932
    },
    {
      "epoch": 2.97,
      "learning_rate": 3.141546898638426e-05,
      "loss": 0.3085,
      "step": 3933
    },
    {
      "epoch": 2.98,
      "learning_rate": 3.1410741301059e-05,
      "loss": 0.2572,
      "step": 3934
    },
    {
      "epoch": 2.98,
      "learning_rate": 3.1406013615733734e-05,
      "loss": 0.2634,
      "step": 3935
    },
    {
      "epoch": 2.98,
      "learning_rate": 3.140128593040847e-05,
      "loss": 0.326,
      "step": 3936
    },
    {
      "epoch": 2.98,
      "learning_rate": 3.1396558245083205e-05,
      "loss": 0.285,
      "step": 3937
    },
    {
      "epoch": 2.98,
      "learning_rate": 3.1391830559757944e-05,
      "loss": 0.2885,
      "step": 3938
    },
    {
      "epoch": 2.98,
      "learning_rate": 3.1387102874432676e-05,
      "loss": 0.2491,
      "step": 3939
    },
    {
      "epoch": 2.98,
      "learning_rate": 3.1382375189107415e-05,
      "loss": 0.2789,
      "step": 3940
    },
    {
      "epoch": 2.98,
      "learning_rate": 3.137764750378215e-05,
      "loss": 0.3107,
      "step": 3941
    },
    {
      "epoch": 2.98,
      "learning_rate": 3.1372919818456886e-05,
      "loss": 0.3228,
      "step": 3942
    },
    {
      "epoch": 2.98,
      "learning_rate": 3.136819213313162e-05,
      "loss": 0.3204,
      "step": 3943
    },
    {
      "epoch": 2.98,
      "learning_rate": 3.136346444780636e-05,
      "loss": 0.2941,
      "step": 3944
    },
    {
      "epoch": 2.98,
      "learning_rate": 3.135873676248109e-05,
      "loss": 0.2544,
      "step": 3945
    },
    {
      "epoch": 2.98,
      "learning_rate": 3.135400907715583e-05,
      "loss": 0.3144,
      "step": 3946
    },
    {
      "epoch": 2.99,
      "learning_rate": 3.134928139183056e-05,
      "loss": 0.3119,
      "step": 3947
    },
    {
      "epoch": 2.99,
      "learning_rate": 3.13445537065053e-05,
      "loss": 0.2462,
      "step": 3948
    },
    {
      "epoch": 2.99,
      "learning_rate": 3.133982602118003e-05,
      "loss": 0.2995,
      "step": 3949
    },
    {
      "epoch": 2.99,
      "learning_rate": 3.133509833585477e-05,
      "loss": 0.3033,
      "step": 3950
    },
    {
      "epoch": 2.99,
      "learning_rate": 3.13303706505295e-05,
      "loss": 0.2512,
      "step": 3951
    },
    {
      "epoch": 2.99,
      "learning_rate": 3.132564296520424e-05,
      "loss": 0.3702,
      "step": 3952
    },
    {
      "epoch": 2.99,
      "learning_rate": 3.132091527987897e-05,
      "loss": 0.2882,
      "step": 3953
    },
    {
      "epoch": 2.99,
      "learning_rate": 3.131618759455371e-05,
      "loss": 0.2686,
      "step": 3954
    },
    {
      "epoch": 2.99,
      "learning_rate": 3.131145990922844e-05,
      "loss": 0.2786,
      "step": 3955
    },
    {
      "epoch": 2.99,
      "learning_rate": 3.130673222390318e-05,
      "loss": 0.2927,
      "step": 3956
    },
    {
      "epoch": 2.99,
      "learning_rate": 3.1302004538577914e-05,
      "loss": 0.2921,
      "step": 3957
    },
    {
      "epoch": 2.99,
      "learning_rate": 3.1297276853252646e-05,
      "loss": 0.2736,
      "step": 3958
    },
    {
      "epoch": 2.99,
      "learning_rate": 3.1292549167927385e-05,
      "loss": 0.2553,
      "step": 3959
    },
    {
      "epoch": 2.99,
      "learning_rate": 3.128782148260212e-05,
      "loss": 0.258,
      "step": 3960
    },
    {
      "epoch": 3.0,
      "learning_rate": 3.128309379727685e-05,
      "loss": 0.2985,
      "step": 3961
    },
    {
      "epoch": 3.0,
      "learning_rate": 3.127836611195159e-05,
      "loss": 0.2827,
      "step": 3962
    },
    {
      "epoch": 3.0,
      "learning_rate": 3.127363842662632e-05,
      "loss": 0.2906,
      "step": 3963
    },
    {
      "epoch": 3.0,
      "learning_rate": 3.126891074130106e-05,
      "loss": 0.2911,
      "step": 3964
    },
    {
      "epoch": 3.0,
      "learning_rate": 3.126418305597579e-05,
      "loss": 0.2498,
      "step": 3965
    },
    {
      "epoch": 3.0,
      "learning_rate": 3.125945537065053e-05,
      "loss": 0.3225,
      "step": 3966
    },
    {
      "epoch": 3.0,
      "learning_rate": 3.125472768532526e-05,
      "loss": 0.2241,
      "step": 3967
    },
    {
      "epoch": 3.0,
      "learning_rate": 3.125e-05,
      "loss": 0.1596,
      "step": 3968
    },
    {
      "epoch": 3.0,
      "learning_rate": 3.124527231467474e-05,
      "loss": 0.125,
      "step": 3969
    },
    {
      "epoch": 3.0,
      "learning_rate": 3.124054462934947e-05,
      "loss": 0.1135,
      "step": 3970
    },
    {
      "epoch": 3.0,
      "learning_rate": 3.123581694402421e-05,
      "loss": 0.1149,
      "step": 3971
    },
    {
      "epoch": 3.0,
      "learning_rate": 3.123108925869894e-05,
      "loss": 0.1435,
      "step": 3972
    },
    {
      "epoch": 3.0,
      "learning_rate": 3.122636157337368e-05,
      "loss": 0.1178,
      "step": 3973
    },
    {
      "epoch": 3.01,
      "learning_rate": 3.122163388804841e-05,
      "loss": 0.1529,
      "step": 3974
    },
    {
      "epoch": 3.01,
      "learning_rate": 3.121690620272315e-05,
      "loss": 0.1819,
      "step": 3975
    },
    {
      "epoch": 3.01,
      "learning_rate": 3.1212178517397884e-05,
      "loss": 0.1347,
      "step": 3976
    },
    {
      "epoch": 3.01,
      "learning_rate": 3.120745083207262e-05,
      "loss": 0.1363,
      "step": 3977
    },
    {
      "epoch": 3.01,
      "learning_rate": 3.1202723146747355e-05,
      "loss": 0.1192,
      "step": 3978
    },
    {
      "epoch": 3.01,
      "learning_rate": 3.1197995461422094e-05,
      "loss": 0.1259,
      "step": 3979
    },
    {
      "epoch": 3.01,
      "learning_rate": 3.1193267776096826e-05,
      "loss": 0.1672,
      "step": 3980
    },
    {
      "epoch": 3.01,
      "learning_rate": 3.1188540090771565e-05,
      "loss": 0.1483,
      "step": 3981
    },
    {
      "epoch": 3.01,
      "learning_rate": 3.11838124054463e-05,
      "loss": 0.1281,
      "step": 3982
    },
    {
      "epoch": 3.01,
      "learning_rate": 3.117908472012103e-05,
      "loss": 0.1283,
      "step": 3983
    },
    {
      "epoch": 3.01,
      "learning_rate": 3.117435703479577e-05,
      "loss": 0.1334,
      "step": 3984
    },
    {
      "epoch": 3.01,
      "learning_rate": 3.11696293494705e-05,
      "loss": 0.1304,
      "step": 3985
    },
    {
      "epoch": 3.01,
      "learning_rate": 3.116490166414523e-05,
      "loss": 0.1258,
      "step": 3986
    },
    {
      "epoch": 3.02,
      "learning_rate": 3.116017397881997e-05,
      "loss": 0.1168,
      "step": 3987
    },
    {
      "epoch": 3.02,
      "learning_rate": 3.11554462934947e-05,
      "loss": 0.1222,
      "step": 3988
    },
    {
      "epoch": 3.02,
      "learning_rate": 3.115071860816944e-05,
      "loss": 0.1614,
      "step": 3989
    },
    {
      "epoch": 3.02,
      "learning_rate": 3.1145990922844174e-05,
      "loss": 0.1417,
      "step": 3990
    },
    {
      "epoch": 3.02,
      "learning_rate": 3.114126323751891e-05,
      "loss": 0.1195,
      "step": 3991
    },
    {
      "epoch": 3.02,
      "learning_rate": 3.1136535552193645e-05,
      "loss": 0.1341,
      "step": 3992
    },
    {
      "epoch": 3.02,
      "learning_rate": 3.1131807866868383e-05,
      "loss": 0.173,
      "step": 3993
    },
    {
      "epoch": 3.02,
      "learning_rate": 3.1127080181543116e-05,
      "loss": 0.1593,
      "step": 3994
    },
    {
      "epoch": 3.02,
      "learning_rate": 3.1122352496217854e-05,
      "loss": 0.1328,
      "step": 3995
    },
    {
      "epoch": 3.02,
      "learning_rate": 3.1117624810892586e-05,
      "loss": 0.1204,
      "step": 3996
    },
    {
      "epoch": 3.02,
      "learning_rate": 3.1112897125567325e-05,
      "loss": 0.1444,
      "step": 3997
    },
    {
      "epoch": 3.02,
      "learning_rate": 3.110816944024206e-05,
      "loss": 0.1369,
      "step": 3998
    },
    {
      "epoch": 3.02,
      "learning_rate": 3.1103441754916796e-05,
      "loss": 0.153,
      "step": 3999
    },
    {
      "epoch": 3.03,
      "learning_rate": 3.109871406959153e-05,
      "loss": 0.1313,
      "step": 4000
    },
    {
      "epoch": 3.03,
      "learning_rate": 3.109398638426627e-05,
      "loss": 0.145,
      "step": 4001
    },
    {
      "epoch": 3.03,
      "learning_rate": 3.1089258698941e-05,
      "loss": 0.1524,
      "step": 4002
    },
    {
      "epoch": 3.03,
      "learning_rate": 3.108453101361574e-05,
      "loss": 0.1493,
      "step": 4003
    },
    {
      "epoch": 3.03,
      "learning_rate": 3.107980332829047e-05,
      "loss": 0.1798,
      "step": 4004
    },
    {
      "epoch": 3.03,
      "learning_rate": 3.107507564296521e-05,
      "loss": 0.1464,
      "step": 4005
    },
    {
      "epoch": 3.03,
      "learning_rate": 3.107034795763994e-05,
      "loss": 0.121,
      "step": 4006
    },
    {
      "epoch": 3.03,
      "learning_rate": 3.106562027231468e-05,
      "loss": 0.1412,
      "step": 4007
    },
    {
      "epoch": 3.03,
      "learning_rate": 3.106089258698941e-05,
      "loss": 0.1482,
      "step": 4008
    },
    {
      "epoch": 3.03,
      "learning_rate": 3.105616490166415e-05,
      "loss": 0.1408,
      "step": 4009
    },
    {
      "epoch": 3.03,
      "learning_rate": 3.105143721633888e-05,
      "loss": 0.1403,
      "step": 4010
    },
    {
      "epoch": 3.03,
      "learning_rate": 3.1046709531013615e-05,
      "loss": 0.141,
      "step": 4011
    },
    {
      "epoch": 3.03,
      "learning_rate": 3.1041981845688354e-05,
      "loss": 0.1362,
      "step": 4012
    },
    {
      "epoch": 3.03,
      "learning_rate": 3.1037254160363086e-05,
      "loss": 0.1258,
      "step": 4013
    },
    {
      "epoch": 3.04,
      "learning_rate": 3.103252647503782e-05,
      "loss": 0.139,
      "step": 4014
    },
    {
      "epoch": 3.04,
      "learning_rate": 3.1027798789712557e-05,
      "loss": 0.1629,
      "step": 4015
    },
    {
      "epoch": 3.04,
      "learning_rate": 3.102307110438729e-05,
      "loss": 0.1366,
      "step": 4016
    },
    {
      "epoch": 3.04,
      "learning_rate": 3.101834341906203e-05,
      "loss": 0.145,
      "step": 4017
    },
    {
      "epoch": 3.04,
      "learning_rate": 3.101361573373676e-05,
      "loss": 0.1228,
      "step": 4018
    },
    {
      "epoch": 3.04,
      "learning_rate": 3.10088880484115e-05,
      "loss": 0.1353,
      "step": 4019
    },
    {
      "epoch": 3.04,
      "learning_rate": 3.100416036308623e-05,
      "loss": 0.1179,
      "step": 4020
    },
    {
      "epoch": 3.04,
      "learning_rate": 3.099943267776097e-05,
      "loss": 0.1462,
      "step": 4021
    },
    {
      "epoch": 3.04,
      "learning_rate": 3.09947049924357e-05,
      "loss": 0.1585,
      "step": 4022
    },
    {
      "epoch": 3.04,
      "learning_rate": 3.098997730711044e-05,
      "loss": 0.1425,
      "step": 4023
    },
    {
      "epoch": 3.04,
      "learning_rate": 3.098524962178517e-05,
      "loss": 0.1392,
      "step": 4024
    },
    {
      "epoch": 3.04,
      "learning_rate": 3.098052193645991e-05,
      "loss": 0.1365,
      "step": 4025
    },
    {
      "epoch": 3.04,
      "learning_rate": 3.097579425113464e-05,
      "loss": 0.1454,
      "step": 4026
    },
    {
      "epoch": 3.05,
      "learning_rate": 3.097106656580938e-05,
      "loss": 0.1553,
      "step": 4027
    },
    {
      "epoch": 3.05,
      "learning_rate": 3.0966338880484114e-05,
      "loss": 0.1386,
      "step": 4028
    },
    {
      "epoch": 3.05,
      "learning_rate": 3.096161119515885e-05,
      "loss": 0.1417,
      "step": 4029
    },
    {
      "epoch": 3.05,
      "learning_rate": 3.095688350983359e-05,
      "loss": 0.1605,
      "step": 4030
    },
    {
      "epoch": 3.05,
      "learning_rate": 3.0952155824508324e-05,
      "loss": 0.1488,
      "step": 4031
    },
    {
      "epoch": 3.05,
      "learning_rate": 3.094742813918306e-05,
      "loss": 0.1305,
      "step": 4032
    },
    {
      "epoch": 3.05,
      "learning_rate": 3.0942700453857795e-05,
      "loss": 0.124,
      "step": 4033
    },
    {
      "epoch": 3.05,
      "learning_rate": 3.0937972768532533e-05,
      "loss": 0.1296,
      "step": 4034
    },
    {
      "epoch": 3.05,
      "learning_rate": 3.0933245083207265e-05,
      "loss": 0.1308,
      "step": 4035
    },
    {
      "epoch": 3.05,
      "learning_rate": 3.0928517397882e-05,
      "loss": 0.1391,
      "step": 4036
    },
    {
      "epoch": 3.05,
      "learning_rate": 3.0923789712556736e-05,
      "loss": 0.1413,
      "step": 4037
    },
    {
      "epoch": 3.05,
      "learning_rate": 3.091906202723147e-05,
      "loss": 0.1326,
      "step": 4038
    },
    {
      "epoch": 3.05,
      "learning_rate": 3.09143343419062e-05,
      "loss": 0.1331,
      "step": 4039
    },
    {
      "epoch": 3.06,
      "learning_rate": 3.090960665658094e-05,
      "loss": 0.1143,
      "step": 4040
    },
    {
      "epoch": 3.06,
      "learning_rate": 3.090487897125567e-05,
      "loss": 0.1395,
      "step": 4041
    },
    {
      "epoch": 3.06,
      "learning_rate": 3.090015128593041e-05,
      "loss": 0.1414,
      "step": 4042
    },
    {
      "epoch": 3.06,
      "learning_rate": 3.089542360060514e-05,
      "loss": 0.143,
      "step": 4043
    },
    {
      "epoch": 3.06,
      "learning_rate": 3.089069591527988e-05,
      "loss": 0.1786,
      "step": 4044
    },
    {
      "epoch": 3.06,
      "learning_rate": 3.088596822995461e-05,
      "loss": 0.1577,
      "step": 4045
    },
    {
      "epoch": 3.06,
      "learning_rate": 3.088124054462935e-05,
      "loss": 0.1461,
      "step": 4046
    },
    {
      "epoch": 3.06,
      "learning_rate": 3.0876512859304084e-05,
      "loss": 0.1565,
      "step": 4047
    },
    {
      "epoch": 3.06,
      "learning_rate": 3.087178517397882e-05,
      "loss": 0.124,
      "step": 4048
    },
    {
      "epoch": 3.06,
      "learning_rate": 3.0867057488653555e-05,
      "loss": 0.1344,
      "step": 4049
    },
    {
      "epoch": 3.06,
      "learning_rate": 3.0862329803328294e-05,
      "loss": 0.1242,
      "step": 4050
    },
    {
      "epoch": 3.06,
      "learning_rate": 3.0857602118003026e-05,
      "loss": 0.1473,
      "step": 4051
    },
    {
      "epoch": 3.06,
      "learning_rate": 3.0852874432677765e-05,
      "loss": 0.1327,
      "step": 4052
    },
    {
      "epoch": 3.07,
      "learning_rate": 3.08481467473525e-05,
      "loss": 0.1278,
      "step": 4053
    },
    {
      "epoch": 3.07,
      "learning_rate": 3.0843419062027236e-05,
      "loss": 0.1881,
      "step": 4054
    },
    {
      "epoch": 3.07,
      "learning_rate": 3.083869137670197e-05,
      "loss": 0.1361,
      "step": 4055
    },
    {
      "epoch": 3.07,
      "learning_rate": 3.0833963691376706e-05,
      "loss": 0.1596,
      "step": 4056
    },
    {
      "epoch": 3.07,
      "learning_rate": 3.082923600605144e-05,
      "loss": 0.1499,
      "step": 4057
    },
    {
      "epoch": 3.07,
      "learning_rate": 3.082450832072618e-05,
      "loss": 0.1542,
      "step": 4058
    },
    {
      "epoch": 3.07,
      "learning_rate": 3.081978063540091e-05,
      "loss": 0.1304,
      "step": 4059
    },
    {
      "epoch": 3.07,
      "learning_rate": 3.081505295007565e-05,
      "loss": 0.1679,
      "step": 4060
    },
    {
      "epoch": 3.07,
      "learning_rate": 3.081032526475038e-05,
      "loss": 0.1403,
      "step": 4061
    },
    {
      "epoch": 3.07,
      "learning_rate": 3.080559757942512e-05,
      "loss": 0.1525,
      "step": 4062
    },
    {
      "epoch": 3.07,
      "learning_rate": 3.080086989409985e-05,
      "loss": 0.1277,
      "step": 4063
    },
    {
      "epoch": 3.07,
      "learning_rate": 3.079614220877458e-05,
      "loss": 0.1583,
      "step": 4064
    },
    {
      "epoch": 3.07,
      "learning_rate": 3.079141452344932e-05,
      "loss": 0.1456,
      "step": 4065
    },
    {
      "epoch": 3.08,
      "learning_rate": 3.0786686838124054e-05,
      "loss": 0.1305,
      "step": 4066
    },
    {
      "epoch": 3.08,
      "learning_rate": 3.0781959152798786e-05,
      "loss": 0.1313,
      "step": 4067
    },
    {
      "epoch": 3.08,
      "learning_rate": 3.0777231467473525e-05,
      "loss": 0.1525,
      "step": 4068
    },
    {
      "epoch": 3.08,
      "learning_rate": 3.077250378214826e-05,
      "loss": 0.143,
      "step": 4069
    },
    {
      "epoch": 3.08,
      "learning_rate": 3.0767776096822996e-05,
      "loss": 0.1411,
      "step": 4070
    },
    {
      "epoch": 3.08,
      "learning_rate": 3.076304841149773e-05,
      "loss": 0.1437,
      "step": 4071
    },
    {
      "epoch": 3.08,
      "learning_rate": 3.075832072617247e-05,
      "loss": 0.1557,
      "step": 4072
    },
    {
      "epoch": 3.08,
      "learning_rate": 3.07535930408472e-05,
      "loss": 0.1527,
      "step": 4073
    },
    {
      "epoch": 3.08,
      "learning_rate": 3.074886535552194e-05,
      "loss": 0.1248,
      "step": 4074
    },
    {
      "epoch": 3.08,
      "learning_rate": 3.074413767019667e-05,
      "loss": 0.1437,
      "step": 4075
    },
    {
      "epoch": 3.08,
      "learning_rate": 3.073940998487141e-05,
      "loss": 0.1454,
      "step": 4076
    },
    {
      "epoch": 3.08,
      "learning_rate": 3.073468229954614e-05,
      "loss": 0.1384,
      "step": 4077
    },
    {
      "epoch": 3.08,
      "learning_rate": 3.072995461422088e-05,
      "loss": 0.1384,
      "step": 4078
    },
    {
      "epoch": 3.08,
      "learning_rate": 3.072522692889561e-05,
      "loss": 0.1457,
      "step": 4079
    },
    {
      "epoch": 3.09,
      "learning_rate": 3.072049924357035e-05,
      "loss": 0.1488,
      "step": 4080
    },
    {
      "epoch": 3.09,
      "learning_rate": 3.071577155824508e-05,
      "loss": 0.1569,
      "step": 4081
    },
    {
      "epoch": 3.09,
      "learning_rate": 3.071104387291982e-05,
      "loss": 0.1542,
      "step": 4082
    },
    {
      "epoch": 3.09,
      "learning_rate": 3.070631618759455e-05,
      "loss": 0.1239,
      "step": 4083
    },
    {
      "epoch": 3.09,
      "learning_rate": 3.070158850226929e-05,
      "loss": 0.1587,
      "step": 4084
    },
    {
      "epoch": 3.09,
      "learning_rate": 3.0696860816944024e-05,
      "loss": 0.1437,
      "step": 4085
    },
    {
      "epoch": 3.09,
      "learning_rate": 3.069213313161876e-05,
      "loss": 0.1494,
      "step": 4086
    },
    {
      "epoch": 3.09,
      "learning_rate": 3.0687405446293495e-05,
      "loss": 0.1436,
      "step": 4087
    },
    {
      "epoch": 3.09,
      "learning_rate": 3.0682677760968234e-05,
      "loss": 0.1531,
      "step": 4088
    },
    {
      "epoch": 3.09,
      "learning_rate": 3.0677950075642966e-05,
      "loss": 0.1274,
      "step": 4089
    },
    {
      "epoch": 3.09,
      "learning_rate": 3.0673222390317705e-05,
      "loss": 0.1545,
      "step": 4090
    },
    {
      "epoch": 3.09,
      "learning_rate": 3.066849470499244e-05,
      "loss": 0.1412,
      "step": 4091
    },
    {
      "epoch": 3.09,
      "learning_rate": 3.066376701966717e-05,
      "loss": 0.1503,
      "step": 4092
    },
    {
      "epoch": 3.1,
      "learning_rate": 3.065903933434191e-05,
      "loss": 0.1337,
      "step": 4093
    },
    {
      "epoch": 3.1,
      "learning_rate": 3.065431164901664e-05,
      "loss": 0.138,
      "step": 4094
    },
    {
      "epoch": 3.1,
      "learning_rate": 3.064958396369138e-05,
      "loss": 0.1283,
      "step": 4095
    },
    {
      "epoch": 3.1,
      "learning_rate": 3.064485627836611e-05,
      "loss": 0.1419,
      "step": 4096
    },
    {
      "epoch": 3.1,
      "learning_rate": 3.064012859304085e-05,
      "loss": 0.1452,
      "step": 4097
    },
    {
      "epoch": 3.1,
      "learning_rate": 3.063540090771558e-05,
      "loss": 0.1615,
      "step": 4098
    },
    {
      "epoch": 3.1,
      "learning_rate": 3.063067322239032e-05,
      "loss": 0.155,
      "step": 4099
    },
    {
      "epoch": 3.1,
      "learning_rate": 3.062594553706505e-05,
      "loss": 0.1473,
      "step": 4100
    },
    {
      "epoch": 3.1,
      "learning_rate": 3.062121785173979e-05,
      "loss": 0.1376,
      "step": 4101
    },
    {
      "epoch": 3.1,
      "learning_rate": 3.0616490166414523e-05,
      "loss": 0.1374,
      "step": 4102
    },
    {
      "epoch": 3.1,
      "learning_rate": 3.061176248108926e-05,
      "loss": 0.1587,
      "step": 4103
    },
    {
      "epoch": 3.1,
      "learning_rate": 3.0607034795763994e-05,
      "loss": 0.1179,
      "step": 4104
    },
    {
      "epoch": 3.1,
      "learning_rate": 3.060230711043873e-05,
      "loss": 0.1748,
      "step": 4105
    },
    {
      "epoch": 3.11,
      "learning_rate": 3.0597579425113465e-05,
      "loss": 0.1546,
      "step": 4106
    },
    {
      "epoch": 3.11,
      "learning_rate": 3.0592851739788204e-05,
      "loss": 0.1381,
      "step": 4107
    },
    {
      "epoch": 3.11,
      "learning_rate": 3.0588124054462936e-05,
      "loss": 0.1566,
      "step": 4108
    },
    {
      "epoch": 3.11,
      "learning_rate": 3.0583396369137675e-05,
      "loss": 0.1767,
      "step": 4109
    },
    {
      "epoch": 3.11,
      "learning_rate": 3.057866868381241e-05,
      "loss": 0.1501,
      "step": 4110
    },
    {
      "epoch": 3.11,
      "learning_rate": 3.0573940998487146e-05,
      "loss": 0.1206,
      "step": 4111
    },
    {
      "epoch": 3.11,
      "learning_rate": 3.056921331316188e-05,
      "loss": 0.1351,
      "step": 4112
    },
    {
      "epoch": 3.11,
      "learning_rate": 3.056448562783662e-05,
      "loss": 0.1429,
      "step": 4113
    },
    {
      "epoch": 3.11,
      "learning_rate": 3.055975794251135e-05,
      "loss": 0.13,
      "step": 4114
    },
    {
      "epoch": 3.11,
      "learning_rate": 3.055503025718609e-05,
      "loss": 0.1105,
      "step": 4115
    },
    {
      "epoch": 3.11,
      "learning_rate": 3.055030257186082e-05,
      "loss": 0.1571,
      "step": 4116
    },
    {
      "epoch": 3.11,
      "learning_rate": 3.054557488653555e-05,
      "loss": 0.1211,
      "step": 4117
    },
    {
      "epoch": 3.11,
      "learning_rate": 3.054084720121029e-05,
      "loss": 0.1324,
      "step": 4118
    },
    {
      "epoch": 3.12,
      "learning_rate": 3.053611951588502e-05,
      "loss": 0.1647,
      "step": 4119
    },
    {
      "epoch": 3.12,
      "learning_rate": 3.0531391830559755e-05,
      "loss": 0.1373,
      "step": 4120
    },
    {
      "epoch": 3.12,
      "learning_rate": 3.0526664145234494e-05,
      "loss": 0.1475,
      "step": 4121
    },
    {
      "epoch": 3.12,
      "learning_rate": 3.0521936459909226e-05,
      "loss": 0.1545,
      "step": 4122
    },
    {
      "epoch": 3.12,
      "learning_rate": 3.0517208774583968e-05,
      "loss": 0.1354,
      "step": 4123
    },
    {
      "epoch": 3.12,
      "learning_rate": 3.05124810892587e-05,
      "loss": 0.1364,
      "step": 4124
    },
    {
      "epoch": 3.12,
      "learning_rate": 3.0507753403933435e-05,
      "loss": 0.1333,
      "step": 4125
    },
    {
      "epoch": 3.12,
      "learning_rate": 3.050302571860817e-05,
      "loss": 0.1326,
      "step": 4126
    },
    {
      "epoch": 3.12,
      "learning_rate": 3.0498298033282906e-05,
      "loss": 0.1319,
      "step": 4127
    },
    {
      "epoch": 3.12,
      "learning_rate": 3.049357034795764e-05,
      "loss": 0.1407,
      "step": 4128
    },
    {
      "epoch": 3.12,
      "learning_rate": 3.0488842662632377e-05,
      "loss": 0.1424,
      "step": 4129
    },
    {
      "epoch": 3.12,
      "learning_rate": 3.048411497730711e-05,
      "loss": 0.139,
      "step": 4130
    },
    {
      "epoch": 3.12,
      "learning_rate": 3.0479387291981848e-05,
      "loss": 0.15,
      "step": 4131
    },
    {
      "epoch": 3.12,
      "learning_rate": 3.047465960665658e-05,
      "loss": 0.1435,
      "step": 4132
    },
    {
      "epoch": 3.13,
      "learning_rate": 3.046993192133132e-05,
      "loss": 0.1352,
      "step": 4133
    },
    {
      "epoch": 3.13,
      "learning_rate": 3.046520423600605e-05,
      "loss": 0.1424,
      "step": 4134
    },
    {
      "epoch": 3.13,
      "learning_rate": 3.046047655068079e-05,
      "loss": 0.1523,
      "step": 4135
    },
    {
      "epoch": 3.13,
      "learning_rate": 3.0455748865355522e-05,
      "loss": 0.1491,
      "step": 4136
    },
    {
      "epoch": 3.13,
      "learning_rate": 3.045102118003026e-05,
      "loss": 0.14,
      "step": 4137
    },
    {
      "epoch": 3.13,
      "learning_rate": 3.0446293494704993e-05,
      "loss": 0.1433,
      "step": 4138
    },
    {
      "epoch": 3.13,
      "learning_rate": 3.0441565809379728e-05,
      "loss": 0.1069,
      "step": 4139
    },
    {
      "epoch": 3.13,
      "learning_rate": 3.043683812405446e-05,
      "loss": 0.1346,
      "step": 4140
    },
    {
      "epoch": 3.13,
      "learning_rate": 3.04321104387292e-05,
      "loss": 0.1662,
      "step": 4141
    },
    {
      "epoch": 3.13,
      "learning_rate": 3.042738275340393e-05,
      "loss": 0.1579,
      "step": 4142
    },
    {
      "epoch": 3.13,
      "learning_rate": 3.042265506807867e-05,
      "loss": 0.1461,
      "step": 4143
    },
    {
      "epoch": 3.13,
      "learning_rate": 3.0417927382753402e-05,
      "loss": 0.1305,
      "step": 4144
    },
    {
      "epoch": 3.13,
      "learning_rate": 3.041319969742814e-05,
      "loss": 0.1608,
      "step": 4145
    },
    {
      "epoch": 3.14,
      "learning_rate": 3.0408472012102873e-05,
      "loss": 0.1259,
      "step": 4146
    },
    {
      "epoch": 3.14,
      "learning_rate": 3.0403744326777612e-05,
      "loss": 0.1681,
      "step": 4147
    },
    {
      "epoch": 3.14,
      "learning_rate": 3.0399016641452344e-05,
      "loss": 0.1521,
      "step": 4148
    },
    {
      "epoch": 3.14,
      "learning_rate": 3.0394288956127083e-05,
      "loss": 0.1315,
      "step": 4149
    },
    {
      "epoch": 3.14,
      "learning_rate": 3.0389561270801818e-05,
      "loss": 0.1583,
      "step": 4150
    },
    {
      "epoch": 3.14,
      "learning_rate": 3.0384833585476554e-05,
      "loss": 0.1541,
      "step": 4151
    },
    {
      "epoch": 3.14,
      "learning_rate": 3.038010590015129e-05,
      "loss": 0.1513,
      "step": 4152
    },
    {
      "epoch": 3.14,
      "learning_rate": 3.037537821482602e-05,
      "loss": 0.1292,
      "step": 4153
    },
    {
      "epoch": 3.14,
      "learning_rate": 3.037065052950076e-05,
      "loss": 0.1486,
      "step": 4154
    },
    {
      "epoch": 3.14,
      "learning_rate": 3.0365922844175492e-05,
      "loss": 0.1383,
      "step": 4155
    },
    {
      "epoch": 3.14,
      "learning_rate": 3.036119515885023e-05,
      "loss": 0.127,
      "step": 4156
    },
    {
      "epoch": 3.14,
      "learning_rate": 3.0356467473524963e-05,
      "loss": 0.1475,
      "step": 4157
    },
    {
      "epoch": 3.14,
      "learning_rate": 3.0351739788199702e-05,
      "loss": 0.1477,
      "step": 4158
    },
    {
      "epoch": 3.15,
      "learning_rate": 3.0347012102874434e-05,
      "loss": 0.1489,
      "step": 4159
    },
    {
      "epoch": 3.15,
      "learning_rate": 3.0342284417549173e-05,
      "loss": 0.131,
      "step": 4160
    },
    {
      "epoch": 3.15,
      "learning_rate": 3.0337556732223905e-05,
      "loss": 0.1503,
      "step": 4161
    },
    {
      "epoch": 3.15,
      "learning_rate": 3.0332829046898644e-05,
      "loss": 0.154,
      "step": 4162
    },
    {
      "epoch": 3.15,
      "learning_rate": 3.0328101361573376e-05,
      "loss": 0.1416,
      "step": 4163
    },
    {
      "epoch": 3.15,
      "learning_rate": 3.032337367624811e-05,
      "loss": 0.1136,
      "step": 4164
    },
    {
      "epoch": 3.15,
      "learning_rate": 3.0318645990922847e-05,
      "loss": 0.174,
      "step": 4165
    },
    {
      "epoch": 3.15,
      "learning_rate": 3.0313918305597582e-05,
      "loss": 0.1484,
      "step": 4166
    },
    {
      "epoch": 3.15,
      "learning_rate": 3.0309190620272314e-05,
      "loss": 0.121,
      "step": 4167
    },
    {
      "epoch": 3.15,
      "learning_rate": 3.0304462934947053e-05,
      "loss": 0.1419,
      "step": 4168
    },
    {
      "epoch": 3.15,
      "learning_rate": 3.0299735249621785e-05,
      "loss": 0.1405,
      "step": 4169
    },
    {
      "epoch": 3.15,
      "learning_rate": 3.0295007564296524e-05,
      "loss": 0.1372,
      "step": 4170
    },
    {
      "epoch": 3.15,
      "learning_rate": 3.0290279878971256e-05,
      "loss": 0.1396,
      "step": 4171
    },
    {
      "epoch": 3.16,
      "learning_rate": 3.0285552193645995e-05,
      "loss": 0.1358,
      "step": 4172
    },
    {
      "epoch": 3.16,
      "learning_rate": 3.0280824508320727e-05,
      "loss": 0.1542,
      "step": 4173
    },
    {
      "epoch": 3.16,
      "learning_rate": 3.0276096822995466e-05,
      "loss": 0.1452,
      "step": 4174
    },
    {
      "epoch": 3.16,
      "learning_rate": 3.0271369137670198e-05,
      "loss": 0.1481,
      "step": 4175
    },
    {
      "epoch": 3.16,
      "learning_rate": 3.0266641452344936e-05,
      "loss": 0.1311,
      "step": 4176
    },
    {
      "epoch": 3.16,
      "learning_rate": 3.026191376701967e-05,
      "loss": 0.141,
      "step": 4177
    },
    {
      "epoch": 3.16,
      "learning_rate": 3.0257186081694404e-05,
      "loss": 0.1489,
      "step": 4178
    },
    {
      "epoch": 3.16,
      "learning_rate": 3.025245839636914e-05,
      "loss": 0.1757,
      "step": 4179
    },
    {
      "epoch": 3.16,
      "learning_rate": 3.0247730711043875e-05,
      "loss": 0.1708,
      "step": 4180
    },
    {
      "epoch": 3.16,
      "learning_rate": 3.0243003025718607e-05,
      "loss": 0.1273,
      "step": 4181
    },
    {
      "epoch": 3.16,
      "learning_rate": 3.0238275340393346e-05,
      "loss": 0.1546,
      "step": 4182
    },
    {
      "epoch": 3.16,
      "learning_rate": 3.0233547655068078e-05,
      "loss": 0.144,
      "step": 4183
    },
    {
      "epoch": 3.16,
      "learning_rate": 3.0228819969742817e-05,
      "loss": 0.1443,
      "step": 4184
    },
    {
      "epoch": 3.17,
      "learning_rate": 3.022409228441755e-05,
      "loss": 0.1584,
      "step": 4185
    },
    {
      "epoch": 3.17,
      "learning_rate": 3.0219364599092288e-05,
      "loss": 0.1625,
      "step": 4186
    },
    {
      "epoch": 3.17,
      "learning_rate": 3.021463691376702e-05,
      "loss": 0.1403,
      "step": 4187
    },
    {
      "epoch": 3.17,
      "learning_rate": 3.020990922844176e-05,
      "loss": 0.1593,
      "step": 4188
    },
    {
      "epoch": 3.17,
      "learning_rate": 3.020518154311649e-05,
      "loss": 0.127,
      "step": 4189
    },
    {
      "epoch": 3.17,
      "learning_rate": 3.020045385779123e-05,
      "loss": 0.1308,
      "step": 4190
    },
    {
      "epoch": 3.17,
      "learning_rate": 3.019572617246596e-05,
      "loss": 0.1339,
      "step": 4191
    },
    {
      "epoch": 3.17,
      "learning_rate": 3.0190998487140697e-05,
      "loss": 0.1564,
      "step": 4192
    },
    {
      "epoch": 3.17,
      "learning_rate": 3.018627080181543e-05,
      "loss": 0.1774,
      "step": 4193
    },
    {
      "epoch": 3.17,
      "learning_rate": 3.0181543116490168e-05,
      "loss": 0.1538,
      "step": 4194
    },
    {
      "epoch": 3.17,
      "learning_rate": 3.01768154311649e-05,
      "loss": 0.1314,
      "step": 4195
    },
    {
      "epoch": 3.17,
      "learning_rate": 3.017208774583964e-05,
      "loss": 0.1555,
      "step": 4196
    },
    {
      "epoch": 3.17,
      "learning_rate": 3.016736006051437e-05,
      "loss": 0.1317,
      "step": 4197
    },
    {
      "epoch": 3.17,
      "learning_rate": 3.016263237518911e-05,
      "loss": 0.1374,
      "step": 4198
    },
    {
      "epoch": 3.18,
      "learning_rate": 3.015790468986384e-05,
      "loss": 0.1374,
      "step": 4199
    },
    {
      "epoch": 3.18,
      "learning_rate": 3.015317700453858e-05,
      "loss": 0.1363,
      "step": 4200
    },
    {
      "epoch": 3.18,
      "learning_rate": 3.0148449319213312e-05,
      "loss": 0.1774,
      "step": 4201
    },
    {
      "epoch": 3.18,
      "learning_rate": 3.014372163388805e-05,
      "loss": 0.1408,
      "step": 4202
    },
    {
      "epoch": 3.18,
      "learning_rate": 3.0138993948562783e-05,
      "loss": 0.1626,
      "step": 4203
    },
    {
      "epoch": 3.18,
      "learning_rate": 3.0134266263237522e-05,
      "loss": 0.1562,
      "step": 4204
    },
    {
      "epoch": 3.18,
      "learning_rate": 3.0129538577912254e-05,
      "loss": 0.1274,
      "step": 4205
    },
    {
      "epoch": 3.18,
      "learning_rate": 3.012481089258699e-05,
      "loss": 0.1324,
      "step": 4206
    },
    {
      "epoch": 3.18,
      "learning_rate": 3.0120083207261722e-05,
      "loss": 0.1857,
      "step": 4207
    },
    {
      "epoch": 3.18,
      "learning_rate": 3.011535552193646e-05,
      "loss": 0.1298,
      "step": 4208
    },
    {
      "epoch": 3.18,
      "learning_rate": 3.01106278366112e-05,
      "loss": 0.134,
      "step": 4209
    },
    {
      "epoch": 3.18,
      "learning_rate": 3.010590015128593e-05,
      "loss": 0.1342,
      "step": 4210
    },
    {
      "epoch": 3.18,
      "learning_rate": 3.010117246596067e-05,
      "loss": 0.1499,
      "step": 4211
    },
    {
      "epoch": 3.19,
      "learning_rate": 3.0096444780635402e-05,
      "loss": 0.1604,
      "step": 4212
    },
    {
      "epoch": 3.19,
      "learning_rate": 3.009171709531014e-05,
      "loss": 0.1556,
      "step": 4213
    },
    {
      "epoch": 3.19,
      "learning_rate": 3.0086989409984873e-05,
      "loss": 0.1224,
      "step": 4214
    },
    {
      "epoch": 3.19,
      "learning_rate": 3.0082261724659612e-05,
      "loss": 0.1527,
      "step": 4215
    },
    {
      "epoch": 3.19,
      "learning_rate": 3.0077534039334344e-05,
      "loss": 0.1353,
      "step": 4216
    },
    {
      "epoch": 3.19,
      "learning_rate": 3.007280635400908e-05,
      "loss": 0.1603,
      "step": 4217
    },
    {
      "epoch": 3.19,
      "learning_rate": 3.0068078668683815e-05,
      "loss": 0.1784,
      "step": 4218
    },
    {
      "epoch": 3.19,
      "learning_rate": 3.006335098335855e-05,
      "loss": 0.1511,
      "step": 4219
    },
    {
      "epoch": 3.19,
      "learning_rate": 3.0058623298033283e-05,
      "loss": 0.1365,
      "step": 4220
    },
    {
      "epoch": 3.19,
      "learning_rate": 3.005389561270802e-05,
      "loss": 0.1487,
      "step": 4221
    },
    {
      "epoch": 3.19,
      "learning_rate": 3.0049167927382753e-05,
      "loss": 0.131,
      "step": 4222
    },
    {
      "epoch": 3.19,
      "learning_rate": 3.0044440242057492e-05,
      "loss": 0.1328,
      "step": 4223
    },
    {
      "epoch": 3.19,
      "learning_rate": 3.0039712556732224e-05,
      "loss": 0.1523,
      "step": 4224
    },
    {
      "epoch": 3.2,
      "learning_rate": 3.0034984871406963e-05,
      "loss": 0.1596,
      "step": 4225
    },
    {
      "epoch": 3.2,
      "learning_rate": 3.0030257186081695e-05,
      "loss": 0.1466,
      "step": 4226
    },
    {
      "epoch": 3.2,
      "learning_rate": 3.0025529500756434e-05,
      "loss": 0.1708,
      "step": 4227
    },
    {
      "epoch": 3.2,
      "learning_rate": 3.0020801815431166e-05,
      "loss": 0.1381,
      "step": 4228
    },
    {
      "epoch": 3.2,
      "learning_rate": 3.0016074130105905e-05,
      "loss": 0.1342,
      "step": 4229
    },
    {
      "epoch": 3.2,
      "learning_rate": 3.0011346444780637e-05,
      "loss": 0.1424,
      "step": 4230
    },
    {
      "epoch": 3.2,
      "learning_rate": 3.0006618759455372e-05,
      "loss": 0.1429,
      "step": 4231
    },
    {
      "epoch": 3.2,
      "learning_rate": 3.0001891074130108e-05,
      "loss": 0.1764,
      "step": 4232
    },
    {
      "epoch": 3.2,
      "learning_rate": 2.9997163388804843e-05,
      "loss": 0.1347,
      "step": 4233
    },
    {
      "epoch": 3.2,
      "learning_rate": 2.9992435703479575e-05,
      "loss": 0.1535,
      "step": 4234
    },
    {
      "epoch": 3.2,
      "learning_rate": 2.9987708018154314e-05,
      "loss": 0.1533,
      "step": 4235
    },
    {
      "epoch": 3.2,
      "learning_rate": 2.9982980332829046e-05,
      "loss": 0.1362,
      "step": 4236
    },
    {
      "epoch": 3.2,
      "learning_rate": 2.9978252647503785e-05,
      "loss": 0.1268,
      "step": 4237
    },
    {
      "epoch": 3.21,
      "learning_rate": 2.9973524962178517e-05,
      "loss": 0.1894,
      "step": 4238
    },
    {
      "epoch": 3.21,
      "learning_rate": 2.9968797276853256e-05,
      "loss": 0.1538,
      "step": 4239
    },
    {
      "epoch": 3.21,
      "learning_rate": 2.9964069591527988e-05,
      "loss": 0.1566,
      "step": 4240
    },
    {
      "epoch": 3.21,
      "learning_rate": 2.9959341906202727e-05,
      "loss": 0.1387,
      "step": 4241
    },
    {
      "epoch": 3.21,
      "learning_rate": 2.995461422087746e-05,
      "loss": 0.1472,
      "step": 4242
    },
    {
      "epoch": 3.21,
      "learning_rate": 2.9949886535552198e-05,
      "loss": 0.1459,
      "step": 4243
    },
    {
      "epoch": 3.21,
      "learning_rate": 2.994515885022693e-05,
      "loss": 0.1483,
      "step": 4244
    },
    {
      "epoch": 3.21,
      "learning_rate": 2.9940431164901665e-05,
      "loss": 0.1514,
      "step": 4245
    },
    {
      "epoch": 3.21,
      "learning_rate": 2.9935703479576397e-05,
      "loss": 0.1361,
      "step": 4246
    },
    {
      "epoch": 3.21,
      "learning_rate": 2.9930975794251136e-05,
      "loss": 0.1449,
      "step": 4247
    },
    {
      "epoch": 3.21,
      "learning_rate": 2.992624810892587e-05,
      "loss": 0.151,
      "step": 4248
    },
    {
      "epoch": 3.21,
      "learning_rate": 2.9921520423600607e-05,
      "loss": 0.123,
      "step": 4249
    },
    {
      "epoch": 3.21,
      "learning_rate": 2.991679273827534e-05,
      "loss": 0.145,
      "step": 4250
    },
    {
      "epoch": 3.21,
      "learning_rate": 2.9912065052950078e-05,
      "loss": 0.1715,
      "step": 4251
    },
    {
      "epoch": 3.22,
      "learning_rate": 2.990733736762481e-05,
      "loss": 0.1858,
      "step": 4252
    },
    {
      "epoch": 3.22,
      "learning_rate": 2.990260968229955e-05,
      "loss": 0.145,
      "step": 4253
    },
    {
      "epoch": 3.22,
      "learning_rate": 2.989788199697428e-05,
      "loss": 0.1479,
      "step": 4254
    },
    {
      "epoch": 3.22,
      "learning_rate": 2.989315431164902e-05,
      "loss": 0.1372,
      "step": 4255
    },
    {
      "epoch": 3.22,
      "learning_rate": 2.9888426626323752e-05,
      "loss": 0.1475,
      "step": 4256
    },
    {
      "epoch": 3.22,
      "learning_rate": 2.988369894099849e-05,
      "loss": 0.1594,
      "step": 4257
    },
    {
      "epoch": 3.22,
      "learning_rate": 2.9878971255673223e-05,
      "loss": 0.1258,
      "step": 4258
    },
    {
      "epoch": 3.22,
      "learning_rate": 2.9874243570347958e-05,
      "loss": 0.1587,
      "step": 4259
    },
    {
      "epoch": 3.22,
      "learning_rate": 2.986951588502269e-05,
      "loss": 0.1756,
      "step": 4260
    },
    {
      "epoch": 3.22,
      "learning_rate": 2.986478819969743e-05,
      "loss": 0.1528,
      "step": 4261
    },
    {
      "epoch": 3.22,
      "learning_rate": 2.986006051437216e-05,
      "loss": 0.1507,
      "step": 4262
    },
    {
      "epoch": 3.22,
      "learning_rate": 2.98553328290469e-05,
      "loss": 0.1382,
      "step": 4263
    },
    {
      "epoch": 3.22,
      "learning_rate": 2.9850605143721632e-05,
      "loss": 0.1649,
      "step": 4264
    },
    {
      "epoch": 3.23,
      "learning_rate": 2.984587745839637e-05,
      "loss": 0.1225,
      "step": 4265
    },
    {
      "epoch": 3.23,
      "learning_rate": 2.9841149773071103e-05,
      "loss": 0.1317,
      "step": 4266
    },
    {
      "epoch": 3.23,
      "learning_rate": 2.9836422087745842e-05,
      "loss": 0.1517,
      "step": 4267
    },
    {
      "epoch": 3.23,
      "learning_rate": 2.9831694402420574e-05,
      "loss": 0.1411,
      "step": 4268
    },
    {
      "epoch": 3.23,
      "learning_rate": 2.9826966717095313e-05,
      "loss": 0.1314,
      "step": 4269
    },
    {
      "epoch": 3.23,
      "learning_rate": 2.9822239031770048e-05,
      "loss": 0.138,
      "step": 4270
    },
    {
      "epoch": 3.23,
      "learning_rate": 2.9817511346444784e-05,
      "loss": 0.1572,
      "step": 4271
    },
    {
      "epoch": 3.23,
      "learning_rate": 2.981278366111952e-05,
      "loss": 0.1481,
      "step": 4272
    },
    {
      "epoch": 3.23,
      "learning_rate": 2.980805597579425e-05,
      "loss": 0.1473,
      "step": 4273
    },
    {
      "epoch": 3.23,
      "learning_rate": 2.980332829046899e-05,
      "loss": 0.1499,
      "step": 4274
    },
    {
      "epoch": 3.23,
      "learning_rate": 2.9798600605143722e-05,
      "loss": 0.1396,
      "step": 4275
    },
    {
      "epoch": 3.23,
      "learning_rate": 2.979387291981846e-05,
      "loss": 0.1433,
      "step": 4276
    },
    {
      "epoch": 3.23,
      "learning_rate": 2.9789145234493193e-05,
      "loss": 0.1414,
      "step": 4277
    },
    {
      "epoch": 3.24,
      "learning_rate": 2.9784417549167932e-05,
      "loss": 0.1408,
      "step": 4278
    },
    {
      "epoch": 3.24,
      "learning_rate": 2.9779689863842664e-05,
      "loss": 0.1432,
      "step": 4279
    },
    {
      "epoch": 3.24,
      "learning_rate": 2.9774962178517403e-05,
      "loss": 0.1361,
      "step": 4280
    },
    {
      "epoch": 3.24,
      "learning_rate": 2.9770234493192135e-05,
      "loss": 0.1296,
      "step": 4281
    },
    {
      "epoch": 3.24,
      "learning_rate": 2.9765506807866874e-05,
      "loss": 0.1674,
      "step": 4282
    },
    {
      "epoch": 3.24,
      "learning_rate": 2.9760779122541606e-05,
      "loss": 0.1215,
      "step": 4283
    },
    {
      "epoch": 3.24,
      "learning_rate": 2.975605143721634e-05,
      "loss": 0.1628,
      "step": 4284
    },
    {
      "epoch": 3.24,
      "learning_rate": 2.9751323751891076e-05,
      "loss": 0.1503,
      "step": 4285
    },
    {
      "epoch": 3.24,
      "learning_rate": 2.9746596066565812e-05,
      "loss": 0.1524,
      "step": 4286
    },
    {
      "epoch": 3.24,
      "learning_rate": 2.9741868381240544e-05,
      "loss": 0.1624,
      "step": 4287
    },
    {
      "epoch": 3.24,
      "learning_rate": 2.9737140695915283e-05,
      "loss": 0.1376,
      "step": 4288
    },
    {
      "epoch": 3.24,
      "learning_rate": 2.9732413010590015e-05,
      "loss": 0.1187,
      "step": 4289
    },
    {
      "epoch": 3.24,
      "learning_rate": 2.9727685325264754e-05,
      "loss": 0.1609,
      "step": 4290
    },
    {
      "epoch": 3.25,
      "learning_rate": 2.9722957639939486e-05,
      "loss": 0.1426,
      "step": 4291
    },
    {
      "epoch": 3.25,
      "learning_rate": 2.9718229954614225e-05,
      "loss": 0.1651,
      "step": 4292
    },
    {
      "epoch": 3.25,
      "learning_rate": 2.9713502269288957e-05,
      "loss": 0.1177,
      "step": 4293
    },
    {
      "epoch": 3.25,
      "learning_rate": 2.9708774583963696e-05,
      "loss": 0.128,
      "step": 4294
    },
    {
      "epoch": 3.25,
      "learning_rate": 2.9704046898638428e-05,
      "loss": 0.1753,
      "step": 4295
    },
    {
      "epoch": 3.25,
      "learning_rate": 2.9699319213313166e-05,
      "loss": 0.1249,
      "step": 4296
    },
    {
      "epoch": 3.25,
      "learning_rate": 2.96945915279879e-05,
      "loss": 0.1454,
      "step": 4297
    },
    {
      "epoch": 3.25,
      "learning_rate": 2.9689863842662634e-05,
      "loss": 0.1422,
      "step": 4298
    },
    {
      "epoch": 3.25,
      "learning_rate": 2.9685136157337366e-05,
      "loss": 0.1498,
      "step": 4299
    },
    {
      "epoch": 3.25,
      "learning_rate": 2.9680408472012105e-05,
      "loss": 0.1687,
      "step": 4300
    },
    {
      "epoch": 3.25,
      "learning_rate": 2.9675680786686837e-05,
      "loss": 0.1264,
      "step": 4301
    },
    {
      "epoch": 3.25,
      "learning_rate": 2.9670953101361576e-05,
      "loss": 0.1607,
      "step": 4302
    },
    {
      "epoch": 3.25,
      "learning_rate": 2.9666225416036308e-05,
      "loss": 0.14,
      "step": 4303
    },
    {
      "epoch": 3.26,
      "learning_rate": 2.9661497730711047e-05,
      "loss": 0.1507,
      "step": 4304
    },
    {
      "epoch": 3.26,
      "learning_rate": 2.965677004538578e-05,
      "loss": 0.1335,
      "step": 4305
    },
    {
      "epoch": 3.26,
      "learning_rate": 2.9652042360060517e-05,
      "loss": 0.1457,
      "step": 4306
    },
    {
      "epoch": 3.26,
      "learning_rate": 2.964731467473525e-05,
      "loss": 0.1375,
      "step": 4307
    },
    {
      "epoch": 3.26,
      "learning_rate": 2.964258698940999e-05,
      "loss": 0.1706,
      "step": 4308
    },
    {
      "epoch": 3.26,
      "learning_rate": 2.963785930408472e-05,
      "loss": 0.1445,
      "step": 4309
    },
    {
      "epoch": 3.26,
      "learning_rate": 2.963313161875946e-05,
      "loss": 0.132,
      "step": 4310
    },
    {
      "epoch": 3.26,
      "learning_rate": 2.962840393343419e-05,
      "loss": 0.1441,
      "step": 4311
    },
    {
      "epoch": 3.26,
      "learning_rate": 2.9623676248108927e-05,
      "loss": 0.1446,
      "step": 4312
    },
    {
      "epoch": 3.26,
      "learning_rate": 2.961894856278366e-05,
      "loss": 0.1706,
      "step": 4313
    },
    {
      "epoch": 3.26,
      "learning_rate": 2.9614220877458398e-05,
      "loss": 0.1419,
      "step": 4314
    },
    {
      "epoch": 3.26,
      "learning_rate": 2.960949319213313e-05,
      "loss": 0.1455,
      "step": 4315
    },
    {
      "epoch": 3.26,
      "learning_rate": 2.960476550680787e-05,
      "loss": 0.1398,
      "step": 4316
    },
    {
      "epoch": 3.26,
      "learning_rate": 2.96000378214826e-05,
      "loss": 0.1615,
      "step": 4317
    },
    {
      "epoch": 3.27,
      "learning_rate": 2.959531013615734e-05,
      "loss": 0.1368,
      "step": 4318
    },
    {
      "epoch": 3.27,
      "learning_rate": 2.959058245083207e-05,
      "loss": 0.1636,
      "step": 4319
    },
    {
      "epoch": 3.27,
      "learning_rate": 2.958585476550681e-05,
      "loss": 0.1658,
      "step": 4320
    },
    {
      "epoch": 3.27,
      "learning_rate": 2.9581127080181542e-05,
      "loss": 0.1318,
      "step": 4321
    },
    {
      "epoch": 3.27,
      "learning_rate": 2.957639939485628e-05,
      "loss": 0.1281,
      "step": 4322
    },
    {
      "epoch": 3.27,
      "learning_rate": 2.9571671709531013e-05,
      "loss": 0.1586,
      "step": 4323
    },
    {
      "epoch": 3.27,
      "learning_rate": 2.9566944024205752e-05,
      "loss": 0.1489,
      "step": 4324
    },
    {
      "epoch": 3.27,
      "learning_rate": 2.9562216338880484e-05,
      "loss": 0.1753,
      "step": 4325
    },
    {
      "epoch": 3.27,
      "learning_rate": 2.955748865355522e-05,
      "loss": 0.141,
      "step": 4326
    },
    {
      "epoch": 3.27,
      "learning_rate": 2.9552760968229952e-05,
      "loss": 0.1523,
      "step": 4327
    },
    {
      "epoch": 3.27,
      "learning_rate": 2.954803328290469e-05,
      "loss": 0.1527,
      "step": 4328
    },
    {
      "epoch": 3.27,
      "learning_rate": 2.954330559757943e-05,
      "loss": 0.1252,
      "step": 4329
    },
    {
      "epoch": 3.27,
      "learning_rate": 2.953857791225416e-05,
      "loss": 0.1246,
      "step": 4330
    },
    {
      "epoch": 3.28,
      "learning_rate": 2.95338502269289e-05,
      "loss": 0.1417,
      "step": 4331
    },
    {
      "epoch": 3.28,
      "learning_rate": 2.9529122541603632e-05,
      "loss": 0.1505,
      "step": 4332
    },
    {
      "epoch": 3.28,
      "learning_rate": 2.952439485627837e-05,
      "loss": 0.1489,
      "step": 4333
    },
    {
      "epoch": 3.28,
      "learning_rate": 2.9519667170953103e-05,
      "loss": 0.1441,
      "step": 4334
    },
    {
      "epoch": 3.28,
      "learning_rate": 2.9514939485627842e-05,
      "loss": 0.1349,
      "step": 4335
    },
    {
      "epoch": 3.28,
      "learning_rate": 2.9510211800302574e-05,
      "loss": 0.1435,
      "step": 4336
    },
    {
      "epoch": 3.28,
      "learning_rate": 2.950548411497731e-05,
      "loss": 0.1521,
      "step": 4337
    },
    {
      "epoch": 3.28,
      "learning_rate": 2.9500756429652045e-05,
      "loss": 0.1433,
      "step": 4338
    },
    {
      "epoch": 3.28,
      "learning_rate": 2.949602874432678e-05,
      "loss": 0.1492,
      "step": 4339
    },
    {
      "epoch": 3.28,
      "learning_rate": 2.9491301059001513e-05,
      "loss": 0.1212,
      "step": 4340
    },
    {
      "epoch": 3.28,
      "learning_rate": 2.948657337367625e-05,
      "loss": 0.1521,
      "step": 4341
    },
    {
      "epoch": 3.28,
      "learning_rate": 2.9481845688350983e-05,
      "loss": 0.1511,
      "step": 4342
    },
    {
      "epoch": 3.28,
      "learning_rate": 2.9477118003025722e-05,
      "loss": 0.157,
      "step": 4343
    },
    {
      "epoch": 3.29,
      "learning_rate": 2.9472390317700454e-05,
      "loss": 0.1286,
      "step": 4344
    },
    {
      "epoch": 3.29,
      "learning_rate": 2.9467662632375193e-05,
      "loss": 0.1492,
      "step": 4345
    },
    {
      "epoch": 3.29,
      "learning_rate": 2.9462934947049925e-05,
      "loss": 0.1407,
      "step": 4346
    },
    {
      "epoch": 3.29,
      "learning_rate": 2.9458207261724664e-05,
      "loss": 0.1336,
      "step": 4347
    },
    {
      "epoch": 3.29,
      "learning_rate": 2.9453479576399396e-05,
      "loss": 0.1449,
      "step": 4348
    },
    {
      "epoch": 3.29,
      "learning_rate": 2.9448751891074135e-05,
      "loss": 0.154,
      "step": 4349
    },
    {
      "epoch": 3.29,
      "learning_rate": 2.9444024205748867e-05,
      "loss": 0.1354,
      "step": 4350
    },
    {
      "epoch": 3.29,
      "learning_rate": 2.9439296520423602e-05,
      "loss": 0.1557,
      "step": 4351
    },
    {
      "epoch": 3.29,
      "learning_rate": 2.9434568835098335e-05,
      "loss": 0.1494,
      "step": 4352
    },
    {
      "epoch": 3.29,
      "learning_rate": 2.9429841149773073e-05,
      "loss": 0.1569,
      "step": 4353
    },
    {
      "epoch": 3.29,
      "learning_rate": 2.9425113464447805e-05,
      "loss": 0.1581,
      "step": 4354
    },
    {
      "epoch": 3.29,
      "learning_rate": 2.9420385779122544e-05,
      "loss": 0.1634,
      "step": 4355
    },
    {
      "epoch": 3.29,
      "learning_rate": 2.9415658093797276e-05,
      "loss": 0.1485,
      "step": 4356
    },
    {
      "epoch": 3.3,
      "learning_rate": 2.9410930408472015e-05,
      "loss": 0.1332,
      "step": 4357
    },
    {
      "epoch": 3.3,
      "learning_rate": 2.9406202723146747e-05,
      "loss": 0.1415,
      "step": 4358
    },
    {
      "epoch": 3.3,
      "learning_rate": 2.9401475037821486e-05,
      "loss": 0.1628,
      "step": 4359
    },
    {
      "epoch": 3.3,
      "learning_rate": 2.9396747352496218e-05,
      "loss": 0.1515,
      "step": 4360
    },
    {
      "epoch": 3.3,
      "learning_rate": 2.9392019667170957e-05,
      "loss": 0.1963,
      "step": 4361
    },
    {
      "epoch": 3.3,
      "learning_rate": 2.938729198184569e-05,
      "loss": 0.1367,
      "step": 4362
    },
    {
      "epoch": 3.3,
      "learning_rate": 2.9382564296520428e-05,
      "loss": 0.1381,
      "step": 4363
    },
    {
      "epoch": 3.3,
      "learning_rate": 2.937783661119516e-05,
      "loss": 0.1377,
      "step": 4364
    },
    {
      "epoch": 3.3,
      "learning_rate": 2.9373108925869895e-05,
      "loss": 0.1453,
      "step": 4365
    },
    {
      "epoch": 3.3,
      "learning_rate": 2.9368381240544627e-05,
      "loss": 0.1556,
      "step": 4366
    },
    {
      "epoch": 3.3,
      "learning_rate": 2.9363653555219366e-05,
      "loss": 0.1625,
      "step": 4367
    },
    {
      "epoch": 3.3,
      "learning_rate": 2.9358925869894098e-05,
      "loss": 0.1355,
      "step": 4368
    },
    {
      "epoch": 3.3,
      "learning_rate": 2.9354198184568837e-05,
      "loss": 0.1423,
      "step": 4369
    },
    {
      "epoch": 3.3,
      "learning_rate": 2.934947049924357e-05,
      "loss": 0.1507,
      "step": 4370
    },
    {
      "epoch": 3.31,
      "learning_rate": 2.9344742813918308e-05,
      "loss": 0.141,
      "step": 4371
    },
    {
      "epoch": 3.31,
      "learning_rate": 2.934001512859304e-05,
      "loss": 0.1629,
      "step": 4372
    },
    {
      "epoch": 3.31,
      "learning_rate": 2.933528744326778e-05,
      "loss": 0.1413,
      "step": 4373
    },
    {
      "epoch": 3.31,
      "learning_rate": 2.933055975794251e-05,
      "loss": 0.1426,
      "step": 4374
    },
    {
      "epoch": 3.31,
      "learning_rate": 2.932583207261725e-05,
      "loss": 0.1548,
      "step": 4375
    },
    {
      "epoch": 3.31,
      "learning_rate": 2.9321104387291982e-05,
      "loss": 0.1381,
      "step": 4376
    },
    {
      "epoch": 3.31,
      "learning_rate": 2.931637670196672e-05,
      "loss": 0.1576,
      "step": 4377
    },
    {
      "epoch": 3.31,
      "learning_rate": 2.9311649016641453e-05,
      "loss": 0.1515,
      "step": 4378
    },
    {
      "epoch": 3.31,
      "learning_rate": 2.9306921331316188e-05,
      "loss": 0.1578,
      "step": 4379
    },
    {
      "epoch": 3.31,
      "learning_rate": 2.930219364599092e-05,
      "loss": 0.1367,
      "step": 4380
    },
    {
      "epoch": 3.31,
      "learning_rate": 2.929746596066566e-05,
      "loss": 0.1474,
      "step": 4381
    },
    {
      "epoch": 3.31,
      "learning_rate": 2.929273827534039e-05,
      "loss": 0.138,
      "step": 4382
    },
    {
      "epoch": 3.31,
      "learning_rate": 2.928801059001513e-05,
      "loss": 0.1316,
      "step": 4383
    },
    {
      "epoch": 3.32,
      "learning_rate": 2.9283282904689862e-05,
      "loss": 0.1183,
      "step": 4384
    },
    {
      "epoch": 3.32,
      "learning_rate": 2.92785552193646e-05,
      "loss": 0.1484,
      "step": 4385
    },
    {
      "epoch": 3.32,
      "learning_rate": 2.9273827534039333e-05,
      "loss": 0.1533,
      "step": 4386
    },
    {
      "epoch": 3.32,
      "learning_rate": 2.9269099848714072e-05,
      "loss": 0.1476,
      "step": 4387
    },
    {
      "epoch": 3.32,
      "learning_rate": 2.9264372163388804e-05,
      "loss": 0.1601,
      "step": 4388
    },
    {
      "epoch": 3.32,
      "learning_rate": 2.9259644478063543e-05,
      "loss": 0.1433,
      "step": 4389
    },
    {
      "epoch": 3.32,
      "learning_rate": 2.9254916792738278e-05,
      "loss": 0.162,
      "step": 4390
    },
    {
      "epoch": 3.32,
      "learning_rate": 2.9250189107413014e-05,
      "loss": 0.1429,
      "step": 4391
    },
    {
      "epoch": 3.32,
      "learning_rate": 2.924546142208775e-05,
      "loss": 0.1449,
      "step": 4392
    },
    {
      "epoch": 3.32,
      "learning_rate": 2.924073373676248e-05,
      "loss": 0.1404,
      "step": 4393
    },
    {
      "epoch": 3.32,
      "learning_rate": 2.923600605143722e-05,
      "loss": 0.1349,
      "step": 4394
    },
    {
      "epoch": 3.32,
      "learning_rate": 2.9231278366111952e-05,
      "loss": 0.1567,
      "step": 4395
    },
    {
      "epoch": 3.32,
      "learning_rate": 2.922655068078669e-05,
      "loss": 0.1844,
      "step": 4396
    },
    {
      "epoch": 3.33,
      "learning_rate": 2.9221822995461423e-05,
      "loss": 0.1449,
      "step": 4397
    },
    {
      "epoch": 3.33,
      "learning_rate": 2.9217095310136162e-05,
      "loss": 0.1399,
      "step": 4398
    },
    {
      "epoch": 3.33,
      "learning_rate": 2.9212367624810894e-05,
      "loss": 0.1626,
      "step": 4399
    },
    {
      "epoch": 3.33,
      "learning_rate": 2.9207639939485633e-05,
      "loss": 0.1331,
      "step": 4400
    },
    {
      "epoch": 3.33,
      "learning_rate": 2.9202912254160365e-05,
      "loss": 0.1299,
      "step": 4401
    },
    {
      "epoch": 3.33,
      "learning_rate": 2.9198184568835103e-05,
      "loss": 0.1566,
      "step": 4402
    },
    {
      "epoch": 3.33,
      "learning_rate": 2.9193456883509836e-05,
      "loss": 0.1492,
      "step": 4403
    },
    {
      "epoch": 3.33,
      "learning_rate": 2.918872919818457e-05,
      "loss": 0.1637,
      "step": 4404
    },
    {
      "epoch": 3.33,
      "learning_rate": 2.9184001512859303e-05,
      "loss": 0.1414,
      "step": 4405
    },
    {
      "epoch": 3.33,
      "learning_rate": 2.9179273827534042e-05,
      "loss": 0.1256,
      "step": 4406
    },
    {
      "epoch": 3.33,
      "learning_rate": 2.9174546142208774e-05,
      "loss": 0.1504,
      "step": 4407
    },
    {
      "epoch": 3.33,
      "learning_rate": 2.9169818456883513e-05,
      "loss": 0.1246,
      "step": 4408
    },
    {
      "epoch": 3.33,
      "learning_rate": 2.9165090771558245e-05,
      "loss": 0.1342,
      "step": 4409
    },
    {
      "epoch": 3.34,
      "learning_rate": 2.9160363086232984e-05,
      "loss": 0.1362,
      "step": 4410
    },
    {
      "epoch": 3.34,
      "learning_rate": 2.9155635400907716e-05,
      "loss": 0.1382,
      "step": 4411
    },
    {
      "epoch": 3.34,
      "learning_rate": 2.9150907715582455e-05,
      "loss": 0.1434,
      "step": 4412
    },
    {
      "epoch": 3.34,
      "learning_rate": 2.9146180030257187e-05,
      "loss": 0.1499,
      "step": 4413
    },
    {
      "epoch": 3.34,
      "learning_rate": 2.9141452344931925e-05,
      "loss": 0.1527,
      "step": 4414
    },
    {
      "epoch": 3.34,
      "learning_rate": 2.9136724659606658e-05,
      "loss": 0.1549,
      "step": 4415
    },
    {
      "epoch": 3.34,
      "learning_rate": 2.9131996974281396e-05,
      "loss": 0.1582,
      "step": 4416
    },
    {
      "epoch": 3.34,
      "learning_rate": 2.912726928895613e-05,
      "loss": 0.1347,
      "step": 4417
    },
    {
      "epoch": 3.34,
      "learning_rate": 2.9122541603630864e-05,
      "loss": 0.141,
      "step": 4418
    },
    {
      "epoch": 3.34,
      "learning_rate": 2.9117813918305596e-05,
      "loss": 0.1486,
      "step": 4419
    },
    {
      "epoch": 3.34,
      "learning_rate": 2.9113086232980335e-05,
      "loss": 0.1537,
      "step": 4420
    },
    {
      "epoch": 3.34,
      "learning_rate": 2.9108358547655067e-05,
      "loss": 0.1707,
      "step": 4421
    },
    {
      "epoch": 3.34,
      "learning_rate": 2.9103630862329806e-05,
      "loss": 0.1363,
      "step": 4422
    },
    {
      "epoch": 3.35,
      "learning_rate": 2.9098903177004538e-05,
      "loss": 0.1409,
      "step": 4423
    },
    {
      "epoch": 3.35,
      "learning_rate": 2.9094175491679277e-05,
      "loss": 0.1447,
      "step": 4424
    },
    {
      "epoch": 3.35,
      "learning_rate": 2.908944780635401e-05,
      "loss": 0.1455,
      "step": 4425
    },
    {
      "epoch": 3.35,
      "learning_rate": 2.9084720121028747e-05,
      "loss": 0.1335,
      "step": 4426
    },
    {
      "epoch": 3.35,
      "learning_rate": 2.907999243570348e-05,
      "loss": 0.1377,
      "step": 4427
    },
    {
      "epoch": 3.35,
      "learning_rate": 2.907526475037822e-05,
      "loss": 0.1542,
      "step": 4428
    },
    {
      "epoch": 3.35,
      "learning_rate": 2.907053706505295e-05,
      "loss": 0.1489,
      "step": 4429
    },
    {
      "epoch": 3.35,
      "learning_rate": 2.906580937972769e-05,
      "loss": 0.1704,
      "step": 4430
    },
    {
      "epoch": 3.35,
      "learning_rate": 2.906108169440242e-05,
      "loss": 0.1457,
      "step": 4431
    },
    {
      "epoch": 3.35,
      "learning_rate": 2.9056354009077157e-05,
      "loss": 0.1379,
      "step": 4432
    },
    {
      "epoch": 3.35,
      "learning_rate": 2.905162632375189e-05,
      "loss": 0.1887,
      "step": 4433
    },
    {
      "epoch": 3.35,
      "learning_rate": 2.9046898638426628e-05,
      "loss": 0.1288,
      "step": 4434
    },
    {
      "epoch": 3.35,
      "learning_rate": 2.904217095310136e-05,
      "loss": 0.144,
      "step": 4435
    },
    {
      "epoch": 3.35,
      "learning_rate": 2.90374432677761e-05,
      "loss": 0.1654,
      "step": 4436
    },
    {
      "epoch": 3.36,
      "learning_rate": 2.903271558245083e-05,
      "loss": 0.1586,
      "step": 4437
    },
    {
      "epoch": 3.36,
      "learning_rate": 2.902798789712557e-05,
      "loss": 0.1422,
      "step": 4438
    },
    {
      "epoch": 3.36,
      "learning_rate": 2.90232602118003e-05,
      "loss": 0.147,
      "step": 4439
    },
    {
      "epoch": 3.36,
      "learning_rate": 2.901853252647504e-05,
      "loss": 0.132,
      "step": 4440
    },
    {
      "epoch": 3.36,
      "learning_rate": 2.9013804841149772e-05,
      "loss": 0.1571,
      "step": 4441
    },
    {
      "epoch": 3.36,
      "learning_rate": 2.900907715582451e-05,
      "loss": 0.1494,
      "step": 4442
    },
    {
      "epoch": 3.36,
      "learning_rate": 2.9004349470499243e-05,
      "loss": 0.1189,
      "step": 4443
    },
    {
      "epoch": 3.36,
      "learning_rate": 2.899962178517398e-05,
      "loss": 0.1351,
      "step": 4444
    },
    {
      "epoch": 3.36,
      "learning_rate": 2.8994894099848714e-05,
      "loss": 0.152,
      "step": 4445
    },
    {
      "epoch": 3.36,
      "learning_rate": 2.899016641452345e-05,
      "loss": 0.1494,
      "step": 4446
    },
    {
      "epoch": 3.36,
      "learning_rate": 2.898543872919818e-05,
      "loss": 0.1385,
      "step": 4447
    },
    {
      "epoch": 3.36,
      "learning_rate": 2.898071104387292e-05,
      "loss": 0.1555,
      "step": 4448
    },
    {
      "epoch": 3.36,
      "learning_rate": 2.897598335854766e-05,
      "loss": 0.1633,
      "step": 4449
    },
    {
      "epoch": 3.37,
      "learning_rate": 2.897125567322239e-05,
      "loss": 0.1478,
      "step": 4450
    },
    {
      "epoch": 3.37,
      "learning_rate": 2.896652798789713e-05,
      "loss": 0.1439,
      "step": 4451
    },
    {
      "epoch": 3.37,
      "learning_rate": 2.8961800302571862e-05,
      "loss": 0.1467,
      "step": 4452
    },
    {
      "epoch": 3.37,
      "learning_rate": 2.89570726172466e-05,
      "loss": 0.1386,
      "step": 4453
    },
    {
      "epoch": 3.37,
      "learning_rate": 2.8952344931921333e-05,
      "loss": 0.1363,
      "step": 4454
    },
    {
      "epoch": 3.37,
      "learning_rate": 2.8947617246596072e-05,
      "loss": 0.138,
      "step": 4455
    },
    {
      "epoch": 3.37,
      "learning_rate": 2.8942889561270804e-05,
      "loss": 0.1347,
      "step": 4456
    },
    {
      "epoch": 3.37,
      "learning_rate": 2.893816187594554e-05,
      "loss": 0.1691,
      "step": 4457
    },
    {
      "epoch": 3.37,
      "learning_rate": 2.893343419062027e-05,
      "loss": 0.1396,
      "step": 4458
    },
    {
      "epoch": 3.37,
      "learning_rate": 2.892870650529501e-05,
      "loss": 0.1479,
      "step": 4459
    },
    {
      "epoch": 3.37,
      "learning_rate": 2.8923978819969743e-05,
      "loss": 0.1417,
      "step": 4460
    },
    {
      "epoch": 3.37,
      "learning_rate": 2.891925113464448e-05,
      "loss": 0.165,
      "step": 4461
    },
    {
      "epoch": 3.37,
      "learning_rate": 2.8914523449319213e-05,
      "loss": 0.143,
      "step": 4462
    },
    {
      "epoch": 3.38,
      "learning_rate": 2.8909795763993952e-05,
      "loss": 0.1545,
      "step": 4463
    },
    {
      "epoch": 3.38,
      "learning_rate": 2.8905068078668684e-05,
      "loss": 0.1616,
      "step": 4464
    },
    {
      "epoch": 3.38,
      "learning_rate": 2.8900340393343423e-05,
      "loss": 0.1487,
      "step": 4465
    },
    {
      "epoch": 3.38,
      "learning_rate": 2.8895612708018155e-05,
      "loss": 0.1585,
      "step": 4466
    },
    {
      "epoch": 3.38,
      "learning_rate": 2.8890885022692894e-05,
      "loss": 0.1516,
      "step": 4467
    },
    {
      "epoch": 3.38,
      "learning_rate": 2.8886157337367626e-05,
      "loss": 0.1681,
      "step": 4468
    },
    {
      "epoch": 3.38,
      "learning_rate": 2.8881429652042365e-05,
      "loss": 0.1302,
      "step": 4469
    },
    {
      "epoch": 3.38,
      "learning_rate": 2.8876701966717097e-05,
      "loss": 0.1497,
      "step": 4470
    },
    {
      "epoch": 3.38,
      "learning_rate": 2.8871974281391832e-05,
      "loss": 0.1989,
      "step": 4471
    },
    {
      "epoch": 3.38,
      "learning_rate": 2.8867246596066564e-05,
      "loss": 0.1407,
      "step": 4472
    },
    {
      "epoch": 3.38,
      "learning_rate": 2.8862518910741303e-05,
      "loss": 0.154,
      "step": 4473
    },
    {
      "epoch": 3.38,
      "learning_rate": 2.8857791225416035e-05,
      "loss": 0.1436,
      "step": 4474
    },
    {
      "epoch": 3.38,
      "learning_rate": 2.8853063540090774e-05,
      "loss": 0.1567,
      "step": 4475
    },
    {
      "epoch": 3.39,
      "learning_rate": 2.8848335854765506e-05,
      "loss": 0.1556,
      "step": 4476
    },
    {
      "epoch": 3.39,
      "learning_rate": 2.8843608169440245e-05,
      "loss": 0.1295,
      "step": 4477
    },
    {
      "epoch": 3.39,
      "learning_rate": 2.8838880484114977e-05,
      "loss": 0.1389,
      "step": 4478
    },
    {
      "epoch": 3.39,
      "learning_rate": 2.8834152798789716e-05,
      "loss": 0.1455,
      "step": 4479
    },
    {
      "epoch": 3.39,
      "learning_rate": 2.8829425113464448e-05,
      "loss": 0.1528,
      "step": 4480
    },
    {
      "epoch": 3.39,
      "learning_rate": 2.8824697428139187e-05,
      "loss": 0.1555,
      "step": 4481
    },
    {
      "epoch": 3.39,
      "learning_rate": 2.881996974281392e-05,
      "loss": 0.1404,
      "step": 4482
    },
    {
      "epoch": 3.39,
      "learning_rate": 2.8815242057488658e-05,
      "loss": 0.1548,
      "step": 4483
    },
    {
      "epoch": 3.39,
      "learning_rate": 2.881051437216339e-05,
      "loss": 0.1343,
      "step": 4484
    },
    {
      "epoch": 3.39,
      "learning_rate": 2.8805786686838125e-05,
      "loss": 0.143,
      "step": 4485
    },
    {
      "epoch": 3.39,
      "learning_rate": 2.8801059001512857e-05,
      "loss": 0.1436,
      "step": 4486
    },
    {
      "epoch": 3.39,
      "learning_rate": 2.8796331316187596e-05,
      "loss": 0.1588,
      "step": 4487
    },
    {
      "epoch": 3.39,
      "learning_rate": 2.8791603630862328e-05,
      "loss": 0.1689,
      "step": 4488
    },
    {
      "epoch": 3.39,
      "learning_rate": 2.8786875945537067e-05,
      "loss": 0.1452,
      "step": 4489
    },
    {
      "epoch": 3.4,
      "learning_rate": 2.87821482602118e-05,
      "loss": 0.1597,
      "step": 4490
    },
    {
      "epoch": 3.4,
      "learning_rate": 2.8777420574886538e-05,
      "loss": 0.1315,
      "step": 4491
    },
    {
      "epoch": 3.4,
      "learning_rate": 2.877269288956127e-05,
      "loss": 0.1499,
      "step": 4492
    },
    {
      "epoch": 3.4,
      "learning_rate": 2.876796520423601e-05,
      "loss": 0.1314,
      "step": 4493
    },
    {
      "epoch": 3.4,
      "learning_rate": 2.876323751891074e-05,
      "loss": 0.1466,
      "step": 4494
    },
    {
      "epoch": 3.4,
      "learning_rate": 2.875850983358548e-05,
      "loss": 0.1386,
      "step": 4495
    },
    {
      "epoch": 3.4,
      "learning_rate": 2.8753782148260212e-05,
      "loss": 0.1486,
      "step": 4496
    },
    {
      "epoch": 3.4,
      "learning_rate": 2.8749054462934947e-05,
      "loss": 0.1376,
      "step": 4497
    },
    {
      "epoch": 3.4,
      "learning_rate": 2.8744326777609683e-05,
      "loss": 0.1317,
      "step": 4498
    },
    {
      "epoch": 3.4,
      "learning_rate": 2.8739599092284418e-05,
      "loss": 0.1554,
      "step": 4499
    },
    {
      "epoch": 3.4,
      "learning_rate": 2.873487140695915e-05,
      "loss": 0.1512,
      "step": 4500
    },
    {
      "epoch": 3.4,
      "learning_rate": 2.873014372163389e-05,
      "loss": 0.1347,
      "step": 4501
    },
    {
      "epoch": 3.4,
      "learning_rate": 2.872541603630862e-05,
      "loss": 0.1457,
      "step": 4502
    },
    {
      "epoch": 3.41,
      "learning_rate": 2.872068835098336e-05,
      "loss": 0.1427,
      "step": 4503
    },
    {
      "epoch": 3.41,
      "learning_rate": 2.8715960665658092e-05,
      "loss": 0.1616,
      "step": 4504
    },
    {
      "epoch": 3.41,
      "learning_rate": 2.871123298033283e-05,
      "loss": 0.1432,
      "step": 4505
    },
    {
      "epoch": 3.41,
      "learning_rate": 2.8706505295007563e-05,
      "loss": 0.1395,
      "step": 4506
    },
    {
      "epoch": 3.41,
      "learning_rate": 2.8701777609682302e-05,
      "loss": 0.1847,
      "step": 4507
    },
    {
      "epoch": 3.41,
      "learning_rate": 2.8697049924357034e-05,
      "loss": 0.1664,
      "step": 4508
    },
    {
      "epoch": 3.41,
      "learning_rate": 2.8692322239031773e-05,
      "loss": 0.1348,
      "step": 4509
    },
    {
      "epoch": 3.41,
      "learning_rate": 2.8687594553706508e-05,
      "loss": 0.1234,
      "step": 4510
    },
    {
      "epoch": 3.41,
      "learning_rate": 2.868286686838124e-05,
      "loss": 0.1598,
      "step": 4511
    },
    {
      "epoch": 3.41,
      "learning_rate": 2.867813918305598e-05,
      "loss": 0.1456,
      "step": 4512
    },
    {
      "epoch": 3.41,
      "learning_rate": 2.867341149773071e-05,
      "loss": 0.1356,
      "step": 4513
    },
    {
      "epoch": 3.41,
      "learning_rate": 2.866868381240545e-05,
      "loss": 0.1411,
      "step": 4514
    },
    {
      "epoch": 3.41,
      "learning_rate": 2.8663956127080182e-05,
      "loss": 0.1453,
      "step": 4515
    },
    {
      "epoch": 3.42,
      "learning_rate": 2.865922844175492e-05,
      "loss": 0.1557,
      "step": 4516
    },
    {
      "epoch": 3.42,
      "learning_rate": 2.8654500756429653e-05,
      "loss": 0.1491,
      "step": 4517
    },
    {
      "epoch": 3.42,
      "learning_rate": 2.864977307110439e-05,
      "loss": 0.1642,
      "step": 4518
    },
    {
      "epoch": 3.42,
      "learning_rate": 2.8645045385779124e-05,
      "loss": 0.148,
      "step": 4519
    },
    {
      "epoch": 3.42,
      "learning_rate": 2.8640317700453863e-05,
      "loss": 0.1293,
      "step": 4520
    },
    {
      "epoch": 3.42,
      "learning_rate": 2.8635590015128595e-05,
      "loss": 0.1394,
      "step": 4521
    },
    {
      "epoch": 3.42,
      "learning_rate": 2.8630862329803333e-05,
      "loss": 0.1566,
      "step": 4522
    },
    {
      "epoch": 3.42,
      "learning_rate": 2.8626134644478066e-05,
      "loss": 0.1432,
      "step": 4523
    },
    {
      "epoch": 3.42,
      "learning_rate": 2.86214069591528e-05,
      "loss": 0.1468,
      "step": 4524
    },
    {
      "epoch": 3.42,
      "learning_rate": 2.8616679273827533e-05,
      "loss": 0.1581,
      "step": 4525
    },
    {
      "epoch": 3.42,
      "learning_rate": 2.8611951588502272e-05,
      "loss": 0.1427,
      "step": 4526
    },
    {
      "epoch": 3.42,
      "learning_rate": 2.8607223903177004e-05,
      "loss": 0.1718,
      "step": 4527
    },
    {
      "epoch": 3.42,
      "learning_rate": 2.8602496217851743e-05,
      "loss": 0.1494,
      "step": 4528
    },
    {
      "epoch": 3.43,
      "learning_rate": 2.8597768532526475e-05,
      "loss": 0.1568,
      "step": 4529
    },
    {
      "epoch": 3.43,
      "learning_rate": 2.8593040847201214e-05,
      "loss": 0.1396,
      "step": 4530
    },
    {
      "epoch": 3.43,
      "learning_rate": 2.8588313161875946e-05,
      "loss": 0.1344,
      "step": 4531
    },
    {
      "epoch": 3.43,
      "learning_rate": 2.8583585476550685e-05,
      "loss": 0.1453,
      "step": 4532
    },
    {
      "epoch": 3.43,
      "learning_rate": 2.8578857791225417e-05,
      "loss": 0.1543,
      "step": 4533
    },
    {
      "epoch": 3.43,
      "learning_rate": 2.8574130105900155e-05,
      "loss": 0.1332,
      "step": 4534
    },
    {
      "epoch": 3.43,
      "learning_rate": 2.8569402420574888e-05,
      "loss": 0.1568,
      "step": 4535
    },
    {
      "epoch": 3.43,
      "learning_rate": 2.8564674735249626e-05,
      "loss": 0.1559,
      "step": 4536
    },
    {
      "epoch": 3.43,
      "learning_rate": 2.855994704992436e-05,
      "loss": 0.1536,
      "step": 4537
    },
    {
      "epoch": 3.43,
      "learning_rate": 2.8555219364599094e-05,
      "loss": 0.1493,
      "step": 4538
    },
    {
      "epoch": 3.43,
      "learning_rate": 2.8550491679273826e-05,
      "loss": 0.1595,
      "step": 4539
    },
    {
      "epoch": 3.43,
      "learning_rate": 2.8545763993948565e-05,
      "loss": 0.1442,
      "step": 4540
    },
    {
      "epoch": 3.43,
      "learning_rate": 2.8541036308623297e-05,
      "loss": 0.1471,
      "step": 4541
    },
    {
      "epoch": 3.44,
      "learning_rate": 2.8536308623298036e-05,
      "loss": 0.1592,
      "step": 4542
    },
    {
      "epoch": 3.44,
      "learning_rate": 2.8531580937972768e-05,
      "loss": 0.1425,
      "step": 4543
    },
    {
      "epoch": 3.44,
      "learning_rate": 2.8526853252647507e-05,
      "loss": 0.1306,
      "step": 4544
    },
    {
      "epoch": 3.44,
      "learning_rate": 2.852212556732224e-05,
      "loss": 0.1445,
      "step": 4545
    },
    {
      "epoch": 3.44,
      "learning_rate": 2.8517397881996977e-05,
      "loss": 0.1272,
      "step": 4546
    },
    {
      "epoch": 3.44,
      "learning_rate": 2.851267019667171e-05,
      "loss": 0.1571,
      "step": 4547
    },
    {
      "epoch": 3.44,
      "learning_rate": 2.850794251134645e-05,
      "loss": 0.1662,
      "step": 4548
    },
    {
      "epoch": 3.44,
      "learning_rate": 2.850321482602118e-05,
      "loss": 0.1529,
      "step": 4549
    },
    {
      "epoch": 3.44,
      "learning_rate": 2.8498487140695916e-05,
      "loss": 0.1446,
      "step": 4550
    },
    {
      "epoch": 3.44,
      "learning_rate": 2.849375945537065e-05,
      "loss": 0.15,
      "step": 4551
    },
    {
      "epoch": 3.44,
      "learning_rate": 2.8489031770045387e-05,
      "loss": 0.1453,
      "step": 4552
    },
    {
      "epoch": 3.44,
      "learning_rate": 2.848430408472012e-05,
      "loss": 0.1424,
      "step": 4553
    },
    {
      "epoch": 3.44,
      "learning_rate": 2.8479576399394858e-05,
      "loss": 0.1572,
      "step": 4554
    },
    {
      "epoch": 3.44,
      "learning_rate": 2.847484871406959e-05,
      "loss": 0.1872,
      "step": 4555
    },
    {
      "epoch": 3.45,
      "learning_rate": 2.847012102874433e-05,
      "loss": 0.1307,
      "step": 4556
    },
    {
      "epoch": 3.45,
      "learning_rate": 2.846539334341906e-05,
      "loss": 0.1655,
      "step": 4557
    },
    {
      "epoch": 3.45,
      "learning_rate": 2.84606656580938e-05,
      "loss": 0.1452,
      "step": 4558
    },
    {
      "epoch": 3.45,
      "learning_rate": 2.845593797276853e-05,
      "loss": 0.152,
      "step": 4559
    },
    {
      "epoch": 3.45,
      "learning_rate": 2.845121028744327e-05,
      "loss": 0.1601,
      "step": 4560
    },
    {
      "epoch": 3.45,
      "learning_rate": 2.8446482602118002e-05,
      "loss": 0.1529,
      "step": 4561
    },
    {
      "epoch": 3.45,
      "learning_rate": 2.844175491679274e-05,
      "loss": 0.1266,
      "step": 4562
    },
    {
      "epoch": 3.45,
      "learning_rate": 2.8437027231467473e-05,
      "loss": 0.1789,
      "step": 4563
    },
    {
      "epoch": 3.45,
      "learning_rate": 2.843229954614221e-05,
      "loss": 0.1243,
      "step": 4564
    },
    {
      "epoch": 3.45,
      "learning_rate": 2.8427571860816944e-05,
      "loss": 0.1466,
      "step": 4565
    },
    {
      "epoch": 3.45,
      "learning_rate": 2.842284417549168e-05,
      "loss": 0.1415,
      "step": 4566
    },
    {
      "epoch": 3.45,
      "learning_rate": 2.841811649016641e-05,
      "loss": 0.1538,
      "step": 4567
    },
    {
      "epoch": 3.45,
      "learning_rate": 2.841338880484115e-05,
      "loss": 0.1349,
      "step": 4568
    },
    {
      "epoch": 3.46,
      "learning_rate": 2.840866111951589e-05,
      "loss": 0.1355,
      "step": 4569
    },
    {
      "epoch": 3.46,
      "learning_rate": 2.840393343419062e-05,
      "loss": 0.1438,
      "step": 4570
    },
    {
      "epoch": 3.46,
      "learning_rate": 2.839920574886536e-05,
      "loss": 0.158,
      "step": 4571
    },
    {
      "epoch": 3.46,
      "learning_rate": 2.8394478063540092e-05,
      "loss": 0.138,
      "step": 4572
    },
    {
      "epoch": 3.46,
      "learning_rate": 2.838975037821483e-05,
      "loss": 0.155,
      "step": 4573
    },
    {
      "epoch": 3.46,
      "learning_rate": 2.8385022692889563e-05,
      "loss": 0.1431,
      "step": 4574
    },
    {
      "epoch": 3.46,
      "learning_rate": 2.8380295007564302e-05,
      "loss": 0.1461,
      "step": 4575
    },
    {
      "epoch": 3.46,
      "learning_rate": 2.8375567322239034e-05,
      "loss": 0.1679,
      "step": 4576
    },
    {
      "epoch": 3.46,
      "learning_rate": 2.837083963691377e-05,
      "loss": 0.1401,
      "step": 4577
    },
    {
      "epoch": 3.46,
      "learning_rate": 2.83661119515885e-05,
      "loss": 0.1612,
      "step": 4578
    },
    {
      "epoch": 3.46,
      "learning_rate": 2.836138426626324e-05,
      "loss": 0.1634,
      "step": 4579
    },
    {
      "epoch": 3.46,
      "learning_rate": 2.8356656580937972e-05,
      "loss": 0.1536,
      "step": 4580
    },
    {
      "epoch": 3.46,
      "learning_rate": 2.835192889561271e-05,
      "loss": 0.154,
      "step": 4581
    },
    {
      "epoch": 3.47,
      "learning_rate": 2.8347201210287443e-05,
      "loss": 0.1424,
      "step": 4582
    },
    {
      "epoch": 3.47,
      "learning_rate": 2.8342473524962182e-05,
      "loss": 0.1503,
      "step": 4583
    },
    {
      "epoch": 3.47,
      "learning_rate": 2.8337745839636914e-05,
      "loss": 0.1422,
      "step": 4584
    },
    {
      "epoch": 3.47,
      "learning_rate": 2.8333018154311653e-05,
      "loss": 0.1437,
      "step": 4585
    },
    {
      "epoch": 3.47,
      "learning_rate": 2.8328290468986385e-05,
      "loss": 0.1446,
      "step": 4586
    },
    {
      "epoch": 3.47,
      "learning_rate": 2.8323562783661124e-05,
      "loss": 0.1498,
      "step": 4587
    },
    {
      "epoch": 3.47,
      "learning_rate": 2.8318835098335856e-05,
      "loss": 0.145,
      "step": 4588
    },
    {
      "epoch": 3.47,
      "learning_rate": 2.8314107413010595e-05,
      "loss": 0.1431,
      "step": 4589
    },
    {
      "epoch": 3.47,
      "learning_rate": 2.8309379727685327e-05,
      "loss": 0.1644,
      "step": 4590
    },
    {
      "epoch": 3.47,
      "learning_rate": 2.8304652042360062e-05,
      "loss": 0.1355,
      "step": 4591
    },
    {
      "epoch": 3.47,
      "learning_rate": 2.8299924357034794e-05,
      "loss": 0.1317,
      "step": 4592
    },
    {
      "epoch": 3.47,
      "learning_rate": 2.8295196671709533e-05,
      "loss": 0.1376,
      "step": 4593
    },
    {
      "epoch": 3.47,
      "learning_rate": 2.8290468986384265e-05,
      "loss": 0.1418,
      "step": 4594
    },
    {
      "epoch": 3.48,
      "learning_rate": 2.8285741301059004e-05,
      "loss": 0.1667,
      "step": 4595
    },
    {
      "epoch": 3.48,
      "learning_rate": 2.8281013615733736e-05,
      "loss": 0.1418,
      "step": 4596
    },
    {
      "epoch": 3.48,
      "learning_rate": 2.8276285930408475e-05,
      "loss": 0.1478,
      "step": 4597
    },
    {
      "epoch": 3.48,
      "learning_rate": 2.8271558245083207e-05,
      "loss": 0.1457,
      "step": 4598
    },
    {
      "epoch": 3.48,
      "learning_rate": 2.8266830559757946e-05,
      "loss": 0.1675,
      "step": 4599
    },
    {
      "epoch": 3.48,
      "learning_rate": 2.8262102874432678e-05,
      "loss": 0.1567,
      "step": 4600
    },
    {
      "epoch": 3.48,
      "learning_rate": 2.8257375189107417e-05,
      "loss": 0.1268,
      "step": 4601
    },
    {
      "epoch": 3.48,
      "learning_rate": 2.825264750378215e-05,
      "loss": 0.1471,
      "step": 4602
    },
    {
      "epoch": 3.48,
      "learning_rate": 2.8247919818456884e-05,
      "loss": 0.1591,
      "step": 4603
    },
    {
      "epoch": 3.48,
      "learning_rate": 2.824319213313162e-05,
      "loss": 0.161,
      "step": 4604
    },
    {
      "epoch": 3.48,
      "learning_rate": 2.8238464447806355e-05,
      "loss": 0.1434,
      "step": 4605
    },
    {
      "epoch": 3.48,
      "learning_rate": 2.8233736762481087e-05,
      "loss": 0.1556,
      "step": 4606
    },
    {
      "epoch": 3.48,
      "learning_rate": 2.8229009077155826e-05,
      "loss": 0.1606,
      "step": 4607
    },
    {
      "epoch": 3.48,
      "learning_rate": 2.8224281391830558e-05,
      "loss": 0.1557,
      "step": 4608
    },
    {
      "epoch": 3.49,
      "learning_rate": 2.8219553706505297e-05,
      "loss": 0.1578,
      "step": 4609
    },
    {
      "epoch": 3.49,
      "learning_rate": 2.821482602118003e-05,
      "loss": 0.14,
      "step": 4610
    },
    {
      "epoch": 3.49,
      "learning_rate": 2.8210098335854768e-05,
      "loss": 0.1275,
      "step": 4611
    },
    {
      "epoch": 3.49,
      "learning_rate": 2.82053706505295e-05,
      "loss": 0.1313,
      "step": 4612
    },
    {
      "epoch": 3.49,
      "learning_rate": 2.820064296520424e-05,
      "loss": 0.1792,
      "step": 4613
    },
    {
      "epoch": 3.49,
      "learning_rate": 2.819591527987897e-05,
      "loss": 0.139,
      "step": 4614
    },
    {
      "epoch": 3.49,
      "learning_rate": 2.819118759455371e-05,
      "loss": 0.1762,
      "step": 4615
    },
    {
      "epoch": 3.49,
      "learning_rate": 2.8186459909228442e-05,
      "loss": 0.1634,
      "step": 4616
    },
    {
      "epoch": 3.49,
      "learning_rate": 2.8181732223903177e-05,
      "loss": 0.1371,
      "step": 4617
    },
    {
      "epoch": 3.49,
      "learning_rate": 2.8177004538577913e-05,
      "loss": 0.1216,
      "step": 4618
    },
    {
      "epoch": 3.49,
      "learning_rate": 2.8172276853252648e-05,
      "loss": 0.1389,
      "step": 4619
    },
    {
      "epoch": 3.49,
      "learning_rate": 2.816754916792738e-05,
      "loss": 0.1235,
      "step": 4620
    },
    {
      "epoch": 3.49,
      "learning_rate": 2.816282148260212e-05,
      "loss": 0.147,
      "step": 4621
    },
    {
      "epoch": 3.5,
      "learning_rate": 2.815809379727685e-05,
      "loss": 0.1588,
      "step": 4622
    },
    {
      "epoch": 3.5,
      "learning_rate": 2.815336611195159e-05,
      "loss": 0.1448,
      "step": 4623
    },
    {
      "epoch": 3.5,
      "learning_rate": 2.8148638426626322e-05,
      "loss": 0.1542,
      "step": 4624
    },
    {
      "epoch": 3.5,
      "learning_rate": 2.814391074130106e-05,
      "loss": 0.146,
      "step": 4625
    },
    {
      "epoch": 3.5,
      "learning_rate": 2.8139183055975793e-05,
      "loss": 0.1493,
      "step": 4626
    },
    {
      "epoch": 3.5,
      "learning_rate": 2.8134455370650532e-05,
      "loss": 0.157,
      "step": 4627
    },
    {
      "epoch": 3.5,
      "learning_rate": 2.8129727685325264e-05,
      "loss": 0.1579,
      "step": 4628
    },
    {
      "epoch": 3.5,
      "learning_rate": 2.8125000000000003e-05,
      "loss": 0.1504,
      "step": 4629
    },
    {
      "epoch": 3.5,
      "learning_rate": 2.8120272314674738e-05,
      "loss": 0.1458,
      "step": 4630
    },
    {
      "epoch": 3.5,
      "learning_rate": 2.811554462934947e-05,
      "loss": 0.1669,
      "step": 4631
    },
    {
      "epoch": 3.5,
      "learning_rate": 2.811081694402421e-05,
      "loss": 0.147,
      "step": 4632
    },
    {
      "epoch": 3.5,
      "learning_rate": 2.810608925869894e-05,
      "loss": 0.155,
      "step": 4633
    },
    {
      "epoch": 3.5,
      "learning_rate": 2.810136157337368e-05,
      "loss": 0.1311,
      "step": 4634
    },
    {
      "epoch": 3.51,
      "learning_rate": 2.8096633888048412e-05,
      "loss": 0.1683,
      "step": 4635
    },
    {
      "epoch": 3.51,
      "learning_rate": 2.809190620272315e-05,
      "loss": 0.1639,
      "step": 4636
    },
    {
      "epoch": 3.51,
      "learning_rate": 2.8087178517397883e-05,
      "loss": 0.1303,
      "step": 4637
    },
    {
      "epoch": 3.51,
      "learning_rate": 2.808245083207262e-05,
      "loss": 0.1568,
      "step": 4638
    },
    {
      "epoch": 3.51,
      "learning_rate": 2.8077723146747354e-05,
      "loss": 0.1426,
      "step": 4639
    },
    {
      "epoch": 3.51,
      "learning_rate": 2.8072995461422093e-05,
      "loss": 0.1352,
      "step": 4640
    },
    {
      "epoch": 3.51,
      "learning_rate": 2.8068267776096825e-05,
      "loss": 0.139,
      "step": 4641
    },
    {
      "epoch": 3.51,
      "learning_rate": 2.8063540090771563e-05,
      "loss": 0.1742,
      "step": 4642
    },
    {
      "epoch": 3.51,
      "learning_rate": 2.8058812405446295e-05,
      "loss": 0.1619,
      "step": 4643
    },
    {
      "epoch": 3.51,
      "learning_rate": 2.805408472012103e-05,
      "loss": 0.1297,
      "step": 4644
    },
    {
      "epoch": 3.51,
      "learning_rate": 2.8049357034795763e-05,
      "loss": 0.1379,
      "step": 4645
    },
    {
      "epoch": 3.51,
      "learning_rate": 2.8044629349470502e-05,
      "loss": 0.1508,
      "step": 4646
    },
    {
      "epoch": 3.51,
      "learning_rate": 2.8039901664145234e-05,
      "loss": 0.1598,
      "step": 4647
    },
    {
      "epoch": 3.52,
      "learning_rate": 2.8035173978819973e-05,
      "loss": 0.1389,
      "step": 4648
    },
    {
      "epoch": 3.52,
      "learning_rate": 2.8030446293494705e-05,
      "loss": 0.1564,
      "step": 4649
    },
    {
      "epoch": 3.52,
      "learning_rate": 2.8025718608169444e-05,
      "loss": 0.1638,
      "step": 4650
    },
    {
      "epoch": 3.52,
      "learning_rate": 2.8020990922844176e-05,
      "loss": 0.1357,
      "step": 4651
    },
    {
      "epoch": 3.52,
      "learning_rate": 2.8016263237518915e-05,
      "loss": 0.1431,
      "step": 4652
    },
    {
      "epoch": 3.52,
      "learning_rate": 2.8011535552193647e-05,
      "loss": 0.1476,
      "step": 4653
    },
    {
      "epoch": 3.52,
      "learning_rate": 2.8006807866868385e-05,
      "loss": 0.1435,
      "step": 4654
    },
    {
      "epoch": 3.52,
      "learning_rate": 2.8002080181543117e-05,
      "loss": 0.1742,
      "step": 4655
    },
    {
      "epoch": 3.52,
      "learning_rate": 2.7997352496217853e-05,
      "loss": 0.1493,
      "step": 4656
    },
    {
      "epoch": 3.52,
      "learning_rate": 2.799262481089259e-05,
      "loss": 0.1476,
      "step": 4657
    },
    {
      "epoch": 3.52,
      "learning_rate": 2.7987897125567324e-05,
      "loss": 0.1584,
      "step": 4658
    },
    {
      "epoch": 3.52,
      "learning_rate": 2.7983169440242056e-05,
      "loss": 0.1654,
      "step": 4659
    },
    {
      "epoch": 3.52,
      "learning_rate": 2.7978441754916795e-05,
      "loss": 0.1378,
      "step": 4660
    },
    {
      "epoch": 3.53,
      "learning_rate": 2.7973714069591527e-05,
      "loss": 0.136,
      "step": 4661
    },
    {
      "epoch": 3.53,
      "learning_rate": 2.7968986384266266e-05,
      "loss": 0.1505,
      "step": 4662
    },
    {
      "epoch": 3.53,
      "learning_rate": 2.7964258698940998e-05,
      "loss": 0.1143,
      "step": 4663
    },
    {
      "epoch": 3.53,
      "learning_rate": 2.7959531013615737e-05,
      "loss": 0.1465,
      "step": 4664
    },
    {
      "epoch": 3.53,
      "learning_rate": 2.795480332829047e-05,
      "loss": 0.1403,
      "step": 4665
    },
    {
      "epoch": 3.53,
      "learning_rate": 2.7950075642965207e-05,
      "loss": 0.1427,
      "step": 4666
    },
    {
      "epoch": 3.53,
      "learning_rate": 2.794534795763994e-05,
      "loss": 0.15,
      "step": 4667
    },
    {
      "epoch": 3.53,
      "learning_rate": 2.7940620272314678e-05,
      "loss": 0.1501,
      "step": 4668
    },
    {
      "epoch": 3.53,
      "learning_rate": 2.793589258698941e-05,
      "loss": 0.129,
      "step": 4669
    },
    {
      "epoch": 3.53,
      "learning_rate": 2.7931164901664146e-05,
      "loss": 0.1492,
      "step": 4670
    },
    {
      "epoch": 3.53,
      "learning_rate": 2.792643721633888e-05,
      "loss": 0.1313,
      "step": 4671
    },
    {
      "epoch": 3.53,
      "learning_rate": 2.7921709531013617e-05,
      "loss": 0.1361,
      "step": 4672
    },
    {
      "epoch": 3.53,
      "learning_rate": 2.791698184568835e-05,
      "loss": 0.1705,
      "step": 4673
    },
    {
      "epoch": 3.53,
      "learning_rate": 2.7912254160363088e-05,
      "loss": 0.14,
      "step": 4674
    },
    {
      "epoch": 3.54,
      "learning_rate": 2.790752647503782e-05,
      "loss": 0.1825,
      "step": 4675
    },
    {
      "epoch": 3.54,
      "learning_rate": 2.790279878971256e-05,
      "loss": 0.1472,
      "step": 4676
    },
    {
      "epoch": 3.54,
      "learning_rate": 2.789807110438729e-05,
      "loss": 0.1413,
      "step": 4677
    },
    {
      "epoch": 3.54,
      "learning_rate": 2.789334341906203e-05,
      "loss": 0.134,
      "step": 4678
    },
    {
      "epoch": 3.54,
      "learning_rate": 2.788861573373676e-05,
      "loss": 0.1541,
      "step": 4679
    },
    {
      "epoch": 3.54,
      "learning_rate": 2.78838880484115e-05,
      "loss": 0.1304,
      "step": 4680
    },
    {
      "epoch": 3.54,
      "learning_rate": 2.7879160363086232e-05,
      "loss": 0.1362,
      "step": 4681
    },
    {
      "epoch": 3.54,
      "learning_rate": 2.787443267776097e-05,
      "loss": 0.1427,
      "step": 4682
    },
    {
      "epoch": 3.54,
      "learning_rate": 2.7869704992435703e-05,
      "loss": 0.1517,
      "step": 4683
    },
    {
      "epoch": 3.54,
      "learning_rate": 2.786497730711044e-05,
      "loss": 0.12,
      "step": 4684
    },
    {
      "epoch": 3.54,
      "learning_rate": 2.7860249621785174e-05,
      "loss": 0.1549,
      "step": 4685
    },
    {
      "epoch": 3.54,
      "learning_rate": 2.785552193645991e-05,
      "loss": 0.1262,
      "step": 4686
    },
    {
      "epoch": 3.54,
      "learning_rate": 2.785079425113464e-05,
      "loss": 0.177,
      "step": 4687
    },
    {
      "epoch": 3.55,
      "learning_rate": 2.784606656580938e-05,
      "loss": 0.1276,
      "step": 4688
    },
    {
      "epoch": 3.55,
      "learning_rate": 2.7841338880484113e-05,
      "loss": 0.1453,
      "step": 4689
    },
    {
      "epoch": 3.55,
      "learning_rate": 2.783661119515885e-05,
      "loss": 0.1599,
      "step": 4690
    },
    {
      "epoch": 3.55,
      "learning_rate": 2.783188350983359e-05,
      "loss": 0.1473,
      "step": 4691
    },
    {
      "epoch": 3.55,
      "learning_rate": 2.7827155824508322e-05,
      "loss": 0.136,
      "step": 4692
    },
    {
      "epoch": 3.55,
      "learning_rate": 2.782242813918306e-05,
      "loss": 0.1522,
      "step": 4693
    },
    {
      "epoch": 3.55,
      "learning_rate": 2.7817700453857793e-05,
      "loss": 0.1572,
      "step": 4694
    },
    {
      "epoch": 3.55,
      "learning_rate": 2.7812972768532532e-05,
      "loss": 0.1518,
      "step": 4695
    },
    {
      "epoch": 3.55,
      "learning_rate": 2.7808245083207264e-05,
      "loss": 0.1401,
      "step": 4696
    },
    {
      "epoch": 3.55,
      "learning_rate": 2.7803517397882e-05,
      "loss": 0.1665,
      "step": 4697
    },
    {
      "epoch": 3.55,
      "learning_rate": 2.779878971255673e-05,
      "loss": 0.1211,
      "step": 4698
    },
    {
      "epoch": 3.55,
      "learning_rate": 2.779406202723147e-05,
      "loss": 0.1502,
      "step": 4699
    },
    {
      "epoch": 3.55,
      "learning_rate": 2.7789334341906202e-05,
      "loss": 0.1331,
      "step": 4700
    },
    {
      "epoch": 3.56,
      "learning_rate": 2.778460665658094e-05,
      "loss": 0.1382,
      "step": 4701
    },
    {
      "epoch": 3.56,
      "learning_rate": 2.7779878971255673e-05,
      "loss": 0.1275,
      "step": 4702
    },
    {
      "epoch": 3.56,
      "learning_rate": 2.7775151285930412e-05,
      "loss": 0.1487,
      "step": 4703
    },
    {
      "epoch": 3.56,
      "learning_rate": 2.7770423600605144e-05,
      "loss": 0.1748,
      "step": 4704
    },
    {
      "epoch": 3.56,
      "learning_rate": 2.7765695915279883e-05,
      "loss": 0.1339,
      "step": 4705
    },
    {
      "epoch": 3.56,
      "learning_rate": 2.7760968229954615e-05,
      "loss": 0.1672,
      "step": 4706
    },
    {
      "epoch": 3.56,
      "learning_rate": 2.7756240544629354e-05,
      "loss": 0.1643,
      "step": 4707
    },
    {
      "epoch": 3.56,
      "learning_rate": 2.7751512859304086e-05,
      "loss": 0.1345,
      "step": 4708
    },
    {
      "epoch": 3.56,
      "learning_rate": 2.774678517397882e-05,
      "loss": 0.1498,
      "step": 4709
    },
    {
      "epoch": 3.56,
      "learning_rate": 2.7742057488653557e-05,
      "loss": 0.1323,
      "step": 4710
    },
    {
      "epoch": 3.56,
      "learning_rate": 2.7737329803328292e-05,
      "loss": 0.1827,
      "step": 4711
    },
    {
      "epoch": 3.56,
      "learning_rate": 2.7732602118003024e-05,
      "loss": 0.1616,
      "step": 4712
    },
    {
      "epoch": 3.56,
      "learning_rate": 2.7727874432677763e-05,
      "loss": 0.161,
      "step": 4713
    },
    {
      "epoch": 3.57,
      "learning_rate": 2.7723146747352495e-05,
      "loss": 0.1356,
      "step": 4714
    },
    {
      "epoch": 3.57,
      "learning_rate": 2.7718419062027234e-05,
      "loss": 0.1397,
      "step": 4715
    },
    {
      "epoch": 3.57,
      "learning_rate": 2.7713691376701966e-05,
      "loss": 0.1325,
      "step": 4716
    },
    {
      "epoch": 3.57,
      "learning_rate": 2.7708963691376705e-05,
      "loss": 0.1502,
      "step": 4717
    },
    {
      "epoch": 3.57,
      "learning_rate": 2.7704236006051437e-05,
      "loss": 0.1636,
      "step": 4718
    },
    {
      "epoch": 3.57,
      "learning_rate": 2.7699508320726176e-05,
      "loss": 0.13,
      "step": 4719
    },
    {
      "epoch": 3.57,
      "learning_rate": 2.7694780635400908e-05,
      "loss": 0.1424,
      "step": 4720
    },
    {
      "epoch": 3.57,
      "learning_rate": 2.7690052950075647e-05,
      "loss": 0.1549,
      "step": 4721
    },
    {
      "epoch": 3.57,
      "learning_rate": 2.768532526475038e-05,
      "loss": 0.1438,
      "step": 4722
    },
    {
      "epoch": 3.57,
      "learning_rate": 2.7680597579425114e-05,
      "loss": 0.1678,
      "step": 4723
    },
    {
      "epoch": 3.57,
      "learning_rate": 2.767586989409985e-05,
      "loss": 0.161,
      "step": 4724
    },
    {
      "epoch": 3.57,
      "learning_rate": 2.7671142208774585e-05,
      "loss": 0.1419,
      "step": 4725
    },
    {
      "epoch": 3.57,
      "learning_rate": 2.7666414523449317e-05,
      "loss": 0.1233,
      "step": 4726
    },
    {
      "epoch": 3.57,
      "learning_rate": 2.7661686838124056e-05,
      "loss": 0.1295,
      "step": 4727
    },
    {
      "epoch": 3.58,
      "learning_rate": 2.7656959152798788e-05,
      "loss": 0.1573,
      "step": 4728
    },
    {
      "epoch": 3.58,
      "learning_rate": 2.7652231467473527e-05,
      "loss": 0.1655,
      "step": 4729
    },
    {
      "epoch": 3.58,
      "learning_rate": 2.764750378214826e-05,
      "loss": 0.1376,
      "step": 4730
    },
    {
      "epoch": 3.58,
      "learning_rate": 2.7642776096822998e-05,
      "loss": 0.1494,
      "step": 4731
    },
    {
      "epoch": 3.58,
      "learning_rate": 2.763804841149773e-05,
      "loss": 0.1519,
      "step": 4732
    },
    {
      "epoch": 3.58,
      "learning_rate": 2.763332072617247e-05,
      "loss": 0.148,
      "step": 4733
    },
    {
      "epoch": 3.58,
      "learning_rate": 2.76285930408472e-05,
      "loss": 0.1454,
      "step": 4734
    },
    {
      "epoch": 3.58,
      "learning_rate": 2.762386535552194e-05,
      "loss": 0.1382,
      "step": 4735
    },
    {
      "epoch": 3.58,
      "learning_rate": 2.7619137670196672e-05,
      "loss": 0.1492,
      "step": 4736
    },
    {
      "epoch": 3.58,
      "learning_rate": 2.7614409984871407e-05,
      "loss": 0.1411,
      "step": 4737
    },
    {
      "epoch": 3.58,
      "learning_rate": 2.7609682299546143e-05,
      "loss": 0.1689,
      "step": 4738
    },
    {
      "epoch": 3.58,
      "learning_rate": 2.7604954614220878e-05,
      "loss": 0.1454,
      "step": 4739
    },
    {
      "epoch": 3.58,
      "learning_rate": 2.760022692889561e-05,
      "loss": 0.1534,
      "step": 4740
    },
    {
      "epoch": 3.59,
      "learning_rate": 2.759549924357035e-05,
      "loss": 0.1501,
      "step": 4741
    },
    {
      "epoch": 3.59,
      "learning_rate": 2.759077155824508e-05,
      "loss": 0.1534,
      "step": 4742
    },
    {
      "epoch": 3.59,
      "learning_rate": 2.758604387291982e-05,
      "loss": 0.1355,
      "step": 4743
    },
    {
      "epoch": 3.59,
      "learning_rate": 2.7581316187594552e-05,
      "loss": 0.1348,
      "step": 4744
    },
    {
      "epoch": 3.59,
      "learning_rate": 2.757658850226929e-05,
      "loss": 0.1497,
      "step": 4745
    },
    {
      "epoch": 3.59,
      "learning_rate": 2.7571860816944023e-05,
      "loss": 0.1487,
      "step": 4746
    },
    {
      "epoch": 3.59,
      "learning_rate": 2.756713313161876e-05,
      "loss": 0.151,
      "step": 4747
    },
    {
      "epoch": 3.59,
      "learning_rate": 2.7562405446293494e-05,
      "loss": 0.168,
      "step": 4748
    },
    {
      "epoch": 3.59,
      "learning_rate": 2.7557677760968233e-05,
      "loss": 0.145,
      "step": 4749
    },
    {
      "epoch": 3.59,
      "learning_rate": 2.7552950075642968e-05,
      "loss": 0.1739,
      "step": 4750
    },
    {
      "epoch": 3.59,
      "learning_rate": 2.75482223903177e-05,
      "loss": 0.1558,
      "step": 4751
    },
    {
      "epoch": 3.59,
      "learning_rate": 2.754349470499244e-05,
      "loss": 0.1608,
      "step": 4752
    },
    {
      "epoch": 3.59,
      "learning_rate": 2.753876701966717e-05,
      "loss": 0.1446,
      "step": 4753
    },
    {
      "epoch": 3.6,
      "learning_rate": 2.753403933434191e-05,
      "loss": 0.1646,
      "step": 4754
    },
    {
      "epoch": 3.6,
      "learning_rate": 2.7529311649016642e-05,
      "loss": 0.1399,
      "step": 4755
    },
    {
      "epoch": 3.6,
      "learning_rate": 2.752458396369138e-05,
      "loss": 0.1414,
      "step": 4756
    },
    {
      "epoch": 3.6,
      "learning_rate": 2.7519856278366113e-05,
      "loss": 0.1822,
      "step": 4757
    },
    {
      "epoch": 3.6,
      "learning_rate": 2.751512859304085e-05,
      "loss": 0.143,
      "step": 4758
    },
    {
      "epoch": 3.6,
      "learning_rate": 2.7510400907715584e-05,
      "loss": 0.1456,
      "step": 4759
    },
    {
      "epoch": 3.6,
      "learning_rate": 2.7505673222390323e-05,
      "loss": 0.1676,
      "step": 4760
    },
    {
      "epoch": 3.6,
      "learning_rate": 2.7500945537065055e-05,
      "loss": 0.1936,
      "step": 4761
    },
    {
      "epoch": 3.6,
      "learning_rate": 2.749621785173979e-05,
      "loss": 0.1521,
      "step": 4762
    },
    {
      "epoch": 3.6,
      "learning_rate": 2.7491490166414525e-05,
      "loss": 0.1773,
      "step": 4763
    },
    {
      "epoch": 3.6,
      "learning_rate": 2.748676248108926e-05,
      "loss": 0.1499,
      "step": 4764
    },
    {
      "epoch": 3.6,
      "learning_rate": 2.7482034795763993e-05,
      "loss": 0.1547,
      "step": 4765
    },
    {
      "epoch": 3.6,
      "learning_rate": 2.7477307110438732e-05,
      "loss": 0.1375,
      "step": 4766
    },
    {
      "epoch": 3.61,
      "learning_rate": 2.7472579425113464e-05,
      "loss": 0.1495,
      "step": 4767
    },
    {
      "epoch": 3.61,
      "learning_rate": 2.7467851739788203e-05,
      "loss": 0.1448,
      "step": 4768
    },
    {
      "epoch": 3.61,
      "learning_rate": 2.7463124054462935e-05,
      "loss": 0.1481,
      "step": 4769
    },
    {
      "epoch": 3.61,
      "learning_rate": 2.7458396369137674e-05,
      "loss": 0.1378,
      "step": 4770
    },
    {
      "epoch": 3.61,
      "learning_rate": 2.7453668683812406e-05,
      "loss": 0.1571,
      "step": 4771
    },
    {
      "epoch": 3.61,
      "learning_rate": 2.7448940998487144e-05,
      "loss": 0.1357,
      "step": 4772
    },
    {
      "epoch": 3.61,
      "learning_rate": 2.7444213313161877e-05,
      "loss": 0.1398,
      "step": 4773
    },
    {
      "epoch": 3.61,
      "learning_rate": 2.7439485627836615e-05,
      "loss": 0.1689,
      "step": 4774
    },
    {
      "epoch": 3.61,
      "learning_rate": 2.7434757942511347e-05,
      "loss": 0.1414,
      "step": 4775
    },
    {
      "epoch": 3.61,
      "learning_rate": 2.7430030257186083e-05,
      "loss": 0.1491,
      "step": 4776
    },
    {
      "epoch": 3.61,
      "learning_rate": 2.742530257186082e-05,
      "loss": 0.1458,
      "step": 4777
    },
    {
      "epoch": 3.61,
      "learning_rate": 2.7420574886535554e-05,
      "loss": 0.1518,
      "step": 4778
    },
    {
      "epoch": 3.61,
      "learning_rate": 2.7415847201210286e-05,
      "loss": 0.1527,
      "step": 4779
    },
    {
      "epoch": 3.62,
      "learning_rate": 2.7411119515885025e-05,
      "loss": 0.1564,
      "step": 4780
    },
    {
      "epoch": 3.62,
      "learning_rate": 2.7406391830559757e-05,
      "loss": 0.1595,
      "step": 4781
    },
    {
      "epoch": 3.62,
      "learning_rate": 2.7401664145234496e-05,
      "loss": 0.124,
      "step": 4782
    },
    {
      "epoch": 3.62,
      "learning_rate": 2.7396936459909228e-05,
      "loss": 0.1463,
      "step": 4783
    },
    {
      "epoch": 3.62,
      "learning_rate": 2.7392208774583966e-05,
      "loss": 0.1366,
      "step": 4784
    },
    {
      "epoch": 3.62,
      "learning_rate": 2.73874810892587e-05,
      "loss": 0.1526,
      "step": 4785
    },
    {
      "epoch": 3.62,
      "learning_rate": 2.7382753403933437e-05,
      "loss": 0.148,
      "step": 4786
    },
    {
      "epoch": 3.62,
      "learning_rate": 2.737802571860817e-05,
      "loss": 0.1637,
      "step": 4787
    },
    {
      "epoch": 3.62,
      "learning_rate": 2.7373298033282908e-05,
      "loss": 0.1533,
      "step": 4788
    },
    {
      "epoch": 3.62,
      "learning_rate": 2.736857034795764e-05,
      "loss": 0.1514,
      "step": 4789
    },
    {
      "epoch": 3.62,
      "learning_rate": 2.7363842662632376e-05,
      "loss": 0.1315,
      "step": 4790
    },
    {
      "epoch": 3.62,
      "learning_rate": 2.735911497730711e-05,
      "loss": 0.1435,
      "step": 4791
    },
    {
      "epoch": 3.62,
      "learning_rate": 2.7354387291981847e-05,
      "loss": 0.1533,
      "step": 4792
    },
    {
      "epoch": 3.62,
      "learning_rate": 2.734965960665658e-05,
      "loss": 0.148,
      "step": 4793
    },
    {
      "epoch": 3.63,
      "learning_rate": 2.7344931921331318e-05,
      "loss": 0.1386,
      "step": 4794
    },
    {
      "epoch": 3.63,
      "learning_rate": 2.734020423600605e-05,
      "loss": 0.1438,
      "step": 4795
    },
    {
      "epoch": 3.63,
      "learning_rate": 2.733547655068079e-05,
      "loss": 0.1346,
      "step": 4796
    },
    {
      "epoch": 3.63,
      "learning_rate": 2.733074886535552e-05,
      "loss": 0.1298,
      "step": 4797
    },
    {
      "epoch": 3.63,
      "learning_rate": 2.732602118003026e-05,
      "loss": 0.1643,
      "step": 4798
    },
    {
      "epoch": 3.63,
      "learning_rate": 2.732129349470499e-05,
      "loss": 0.1446,
      "step": 4799
    },
    {
      "epoch": 3.63,
      "learning_rate": 2.731656580937973e-05,
      "loss": 0.1316,
      "step": 4800
    },
    {
      "epoch": 3.63,
      "learning_rate": 2.7311838124054462e-05,
      "loss": 0.1426,
      "step": 4801
    },
    {
      "epoch": 3.63,
      "learning_rate": 2.73071104387292e-05,
      "loss": 0.1639,
      "step": 4802
    },
    {
      "epoch": 3.63,
      "learning_rate": 2.7302382753403933e-05,
      "loss": 0.1551,
      "step": 4803
    },
    {
      "epoch": 3.63,
      "learning_rate": 2.729765506807867e-05,
      "loss": 0.1393,
      "step": 4804
    },
    {
      "epoch": 3.63,
      "learning_rate": 2.7292927382753404e-05,
      "loss": 0.154,
      "step": 4805
    },
    {
      "epoch": 3.63,
      "learning_rate": 2.728819969742814e-05,
      "loss": 0.1676,
      "step": 4806
    },
    {
      "epoch": 3.64,
      "learning_rate": 2.728347201210287e-05,
      "loss": 0.137,
      "step": 4807
    },
    {
      "epoch": 3.64,
      "learning_rate": 2.727874432677761e-05,
      "loss": 0.1428,
      "step": 4808
    },
    {
      "epoch": 3.64,
      "learning_rate": 2.7274016641452342e-05,
      "loss": 0.1261,
      "step": 4809
    },
    {
      "epoch": 3.64,
      "learning_rate": 2.726928895612708e-05,
      "loss": 0.1552,
      "step": 4810
    },
    {
      "epoch": 3.64,
      "learning_rate": 2.726456127080182e-05,
      "loss": 0.1432,
      "step": 4811
    },
    {
      "epoch": 3.64,
      "learning_rate": 2.7259833585476552e-05,
      "loss": 0.162,
      "step": 4812
    },
    {
      "epoch": 3.64,
      "learning_rate": 2.725510590015129e-05,
      "loss": 0.1195,
      "step": 4813
    },
    {
      "epoch": 3.64,
      "learning_rate": 2.7250378214826023e-05,
      "loss": 0.1546,
      "step": 4814
    },
    {
      "epoch": 3.64,
      "learning_rate": 2.724565052950076e-05,
      "loss": 0.1356,
      "step": 4815
    },
    {
      "epoch": 3.64,
      "learning_rate": 2.7240922844175494e-05,
      "loss": 0.1387,
      "step": 4816
    },
    {
      "epoch": 3.64,
      "learning_rate": 2.723619515885023e-05,
      "loss": 0.1342,
      "step": 4817
    },
    {
      "epoch": 3.64,
      "learning_rate": 2.723146747352496e-05,
      "loss": 0.1348,
      "step": 4818
    },
    {
      "epoch": 3.64,
      "learning_rate": 2.72267397881997e-05,
      "loss": 0.1467,
      "step": 4819
    },
    {
      "epoch": 3.65,
      "learning_rate": 2.7222012102874432e-05,
      "loss": 0.1384,
      "step": 4820
    },
    {
      "epoch": 3.65,
      "learning_rate": 2.721728441754917e-05,
      "loss": 0.152,
      "step": 4821
    },
    {
      "epoch": 3.65,
      "learning_rate": 2.7212556732223903e-05,
      "loss": 0.1835,
      "step": 4822
    },
    {
      "epoch": 3.65,
      "learning_rate": 2.7207829046898642e-05,
      "loss": 0.1411,
      "step": 4823
    },
    {
      "epoch": 3.65,
      "learning_rate": 2.7203101361573374e-05,
      "loss": 0.1333,
      "step": 4824
    },
    {
      "epoch": 3.65,
      "learning_rate": 2.7198373676248113e-05,
      "loss": 0.1788,
      "step": 4825
    },
    {
      "epoch": 3.65,
      "learning_rate": 2.7193645990922845e-05,
      "loss": 0.1356,
      "step": 4826
    },
    {
      "epoch": 3.65,
      "learning_rate": 2.7188918305597584e-05,
      "loss": 0.1313,
      "step": 4827
    },
    {
      "epoch": 3.65,
      "learning_rate": 2.7184190620272316e-05,
      "loss": 0.1565,
      "step": 4828
    },
    {
      "epoch": 3.65,
      "learning_rate": 2.717946293494705e-05,
      "loss": 0.1581,
      "step": 4829
    },
    {
      "epoch": 3.65,
      "learning_rate": 2.7174735249621787e-05,
      "loss": 0.1884,
      "step": 4830
    },
    {
      "epoch": 3.65,
      "learning_rate": 2.7170007564296522e-05,
      "loss": 0.122,
      "step": 4831
    },
    {
      "epoch": 3.65,
      "learning_rate": 2.7165279878971254e-05,
      "loss": 0.1596,
      "step": 4832
    },
    {
      "epoch": 3.66,
      "learning_rate": 2.7160552193645993e-05,
      "loss": 0.155,
      "step": 4833
    },
    {
      "epoch": 3.66,
      "learning_rate": 2.7155824508320725e-05,
      "loss": 0.1572,
      "step": 4834
    },
    {
      "epoch": 3.66,
      "learning_rate": 2.7151096822995464e-05,
      "loss": 0.1534,
      "step": 4835
    },
    {
      "epoch": 3.66,
      "learning_rate": 2.7146369137670196e-05,
      "loss": 0.1547,
      "step": 4836
    },
    {
      "epoch": 3.66,
      "learning_rate": 2.7141641452344935e-05,
      "loss": 0.1536,
      "step": 4837
    },
    {
      "epoch": 3.66,
      "learning_rate": 2.7136913767019667e-05,
      "loss": 0.136,
      "step": 4838
    },
    {
      "epoch": 3.66,
      "learning_rate": 2.7132186081694406e-05,
      "loss": 0.1349,
      "step": 4839
    },
    {
      "epoch": 3.66,
      "learning_rate": 2.7127458396369138e-05,
      "loss": 0.1692,
      "step": 4840
    },
    {
      "epoch": 3.66,
      "learning_rate": 2.7122730711043877e-05,
      "loss": 0.139,
      "step": 4841
    },
    {
      "epoch": 3.66,
      "learning_rate": 2.711800302571861e-05,
      "loss": 0.1356,
      "step": 4842
    },
    {
      "epoch": 3.66,
      "learning_rate": 2.7113275340393344e-05,
      "loss": 0.1379,
      "step": 4843
    },
    {
      "epoch": 3.66,
      "learning_rate": 2.710854765506808e-05,
      "loss": 0.145,
      "step": 4844
    },
    {
      "epoch": 3.66,
      "learning_rate": 2.7103819969742815e-05,
      "loss": 0.1412,
      "step": 4845
    },
    {
      "epoch": 3.66,
      "learning_rate": 2.7099092284417547e-05,
      "loss": 0.1519,
      "step": 4846
    },
    {
      "epoch": 3.67,
      "learning_rate": 2.7094364599092286e-05,
      "loss": 0.1448,
      "step": 4847
    },
    {
      "epoch": 3.67,
      "learning_rate": 2.7089636913767018e-05,
      "loss": 0.1662,
      "step": 4848
    },
    {
      "epoch": 3.67,
      "learning_rate": 2.7084909228441757e-05,
      "loss": 0.1338,
      "step": 4849
    },
    {
      "epoch": 3.67,
      "learning_rate": 2.708018154311649e-05,
      "loss": 0.1421,
      "step": 4850
    },
    {
      "epoch": 3.67,
      "learning_rate": 2.7075453857791228e-05,
      "loss": 0.1449,
      "step": 4851
    },
    {
      "epoch": 3.67,
      "learning_rate": 2.707072617246596e-05,
      "loss": 0.1364,
      "step": 4852
    },
    {
      "epoch": 3.67,
      "learning_rate": 2.70659984871407e-05,
      "loss": 0.1308,
      "step": 4853
    },
    {
      "epoch": 3.67,
      "learning_rate": 2.706127080181543e-05,
      "loss": 0.1534,
      "step": 4854
    },
    {
      "epoch": 3.67,
      "learning_rate": 2.705654311649017e-05,
      "loss": 0.1384,
      "step": 4855
    },
    {
      "epoch": 3.67,
      "learning_rate": 2.7051815431164902e-05,
      "loss": 0.1626,
      "step": 4856
    },
    {
      "epoch": 3.67,
      "learning_rate": 2.7047087745839637e-05,
      "loss": 0.1415,
      "step": 4857
    },
    {
      "epoch": 3.67,
      "learning_rate": 2.7042360060514373e-05,
      "loss": 0.1609,
      "step": 4858
    },
    {
      "epoch": 3.67,
      "learning_rate": 2.7037632375189108e-05,
      "loss": 0.1555,
      "step": 4859
    },
    {
      "epoch": 3.68,
      "learning_rate": 2.703290468986384e-05,
      "loss": 0.1712,
      "step": 4860
    },
    {
      "epoch": 3.68,
      "learning_rate": 2.702817700453858e-05,
      "loss": 0.1426,
      "step": 4861
    },
    {
      "epoch": 3.68,
      "learning_rate": 2.702344931921331e-05,
      "loss": 0.1407,
      "step": 4862
    },
    {
      "epoch": 3.68,
      "learning_rate": 2.701872163388805e-05,
      "loss": 0.1464,
      "step": 4863
    },
    {
      "epoch": 3.68,
      "learning_rate": 2.7013993948562782e-05,
      "loss": 0.1427,
      "step": 4864
    },
    {
      "epoch": 3.68,
      "learning_rate": 2.700926626323752e-05,
      "loss": 0.1489,
      "step": 4865
    },
    {
      "epoch": 3.68,
      "learning_rate": 2.7004538577912253e-05,
      "loss": 0.1546,
      "step": 4866
    },
    {
      "epoch": 3.68,
      "learning_rate": 2.699981089258699e-05,
      "loss": 0.1703,
      "step": 4867
    },
    {
      "epoch": 3.68,
      "learning_rate": 2.6995083207261724e-05,
      "loss": 0.1451,
      "step": 4868
    },
    {
      "epoch": 3.68,
      "learning_rate": 2.6990355521936463e-05,
      "loss": 0.1525,
      "step": 4869
    },
    {
      "epoch": 3.68,
      "learning_rate": 2.6985627836611198e-05,
      "loss": 0.145,
      "step": 4870
    },
    {
      "epoch": 3.68,
      "learning_rate": 2.698090015128593e-05,
      "loss": 0.1507,
      "step": 4871
    },
    {
      "epoch": 3.68,
      "learning_rate": 2.697617246596067e-05,
      "loss": 0.1381,
      "step": 4872
    },
    {
      "epoch": 3.69,
      "learning_rate": 2.69714447806354e-05,
      "loss": 0.1485,
      "step": 4873
    },
    {
      "epoch": 3.69,
      "learning_rate": 2.696671709531014e-05,
      "loss": 0.1476,
      "step": 4874
    },
    {
      "epoch": 3.69,
      "learning_rate": 2.6961989409984872e-05,
      "loss": 0.1378,
      "step": 4875
    },
    {
      "epoch": 3.69,
      "learning_rate": 2.695726172465961e-05,
      "loss": 0.1348,
      "step": 4876
    },
    {
      "epoch": 3.69,
      "learning_rate": 2.6952534039334343e-05,
      "loss": 0.1428,
      "step": 4877
    },
    {
      "epoch": 3.69,
      "learning_rate": 2.694780635400908e-05,
      "loss": 0.1201,
      "step": 4878
    },
    {
      "epoch": 3.69,
      "learning_rate": 2.6943078668683814e-05,
      "loss": 0.1665,
      "step": 4879
    },
    {
      "epoch": 3.69,
      "learning_rate": 2.6938350983358552e-05,
      "loss": 0.149,
      "step": 4880
    },
    {
      "epoch": 3.69,
      "learning_rate": 2.6933623298033285e-05,
      "loss": 0.1612,
      "step": 4881
    },
    {
      "epoch": 3.69,
      "learning_rate": 2.692889561270802e-05,
      "loss": 0.1372,
      "step": 4882
    },
    {
      "epoch": 3.69,
      "learning_rate": 2.6924167927382755e-05,
      "loss": 0.1419,
      "step": 4883
    },
    {
      "epoch": 3.69,
      "learning_rate": 2.691944024205749e-05,
      "loss": 0.1311,
      "step": 4884
    },
    {
      "epoch": 3.69,
      "learning_rate": 2.6914712556732223e-05,
      "loss": 0.1598,
      "step": 4885
    },
    {
      "epoch": 3.7,
      "learning_rate": 2.6909984871406962e-05,
      "loss": 0.1586,
      "step": 4886
    },
    {
      "epoch": 3.7,
      "learning_rate": 2.6905257186081694e-05,
      "loss": 0.146,
      "step": 4887
    },
    {
      "epoch": 3.7,
      "learning_rate": 2.6900529500756433e-05,
      "loss": 0.1295,
      "step": 4888
    },
    {
      "epoch": 3.7,
      "learning_rate": 2.6895801815431165e-05,
      "loss": 0.1635,
      "step": 4889
    },
    {
      "epoch": 3.7,
      "learning_rate": 2.6891074130105904e-05,
      "loss": 0.166,
      "step": 4890
    },
    {
      "epoch": 3.7,
      "learning_rate": 2.6886346444780636e-05,
      "loss": 0.1735,
      "step": 4891
    },
    {
      "epoch": 3.7,
      "learning_rate": 2.6881618759455374e-05,
      "loss": 0.1678,
      "step": 4892
    },
    {
      "epoch": 3.7,
      "learning_rate": 2.6876891074130107e-05,
      "loss": 0.1611,
      "step": 4893
    },
    {
      "epoch": 3.7,
      "learning_rate": 2.6872163388804845e-05,
      "loss": 0.1364,
      "step": 4894
    },
    {
      "epoch": 3.7,
      "learning_rate": 2.6867435703479577e-05,
      "loss": 0.1541,
      "step": 4895
    },
    {
      "epoch": 3.7,
      "learning_rate": 2.6862708018154313e-05,
      "loss": 0.1745,
      "step": 4896
    },
    {
      "epoch": 3.7,
      "learning_rate": 2.6857980332829048e-05,
      "loss": 0.1353,
      "step": 4897
    },
    {
      "epoch": 3.7,
      "learning_rate": 2.6853252647503784e-05,
      "loss": 0.1483,
      "step": 4898
    },
    {
      "epoch": 3.71,
      "learning_rate": 2.6848524962178516e-05,
      "loss": 0.1693,
      "step": 4899
    },
    {
      "epoch": 3.71,
      "learning_rate": 2.6843797276853255e-05,
      "loss": 0.1564,
      "step": 4900
    },
    {
      "epoch": 3.71,
      "learning_rate": 2.6839069591527987e-05,
      "loss": 0.1339,
      "step": 4901
    },
    {
      "epoch": 3.71,
      "learning_rate": 2.6834341906202726e-05,
      "loss": 0.138,
      "step": 4902
    },
    {
      "epoch": 3.71,
      "learning_rate": 2.6829614220877458e-05,
      "loss": 0.1571,
      "step": 4903
    },
    {
      "epoch": 3.71,
      "learning_rate": 2.6824886535552196e-05,
      "loss": 0.1414,
      "step": 4904
    },
    {
      "epoch": 3.71,
      "learning_rate": 2.682015885022693e-05,
      "loss": 0.1465,
      "step": 4905
    },
    {
      "epoch": 3.71,
      "learning_rate": 2.6815431164901667e-05,
      "loss": 0.1349,
      "step": 4906
    },
    {
      "epoch": 3.71,
      "learning_rate": 2.68107034795764e-05,
      "loss": 0.1557,
      "step": 4907
    },
    {
      "epoch": 3.71,
      "learning_rate": 2.6805975794251138e-05,
      "loss": 0.1344,
      "step": 4908
    },
    {
      "epoch": 3.71,
      "learning_rate": 2.680124810892587e-05,
      "loss": 0.1476,
      "step": 4909
    },
    {
      "epoch": 3.71,
      "learning_rate": 2.6796520423600606e-05,
      "loss": 0.1836,
      "step": 4910
    },
    {
      "epoch": 3.71,
      "learning_rate": 2.679179273827534e-05,
      "loss": 0.1381,
      "step": 4911
    },
    {
      "epoch": 3.71,
      "learning_rate": 2.6787065052950077e-05,
      "loss": 0.1381,
      "step": 4912
    },
    {
      "epoch": 3.72,
      "learning_rate": 2.678233736762481e-05,
      "loss": 0.1404,
      "step": 4913
    },
    {
      "epoch": 3.72,
      "learning_rate": 2.6777609682299548e-05,
      "loss": 0.1638,
      "step": 4914
    },
    {
      "epoch": 3.72,
      "learning_rate": 2.677288199697428e-05,
      "loss": 0.1825,
      "step": 4915
    },
    {
      "epoch": 3.72,
      "learning_rate": 2.676815431164902e-05,
      "loss": 0.1844,
      "step": 4916
    },
    {
      "epoch": 3.72,
      "learning_rate": 2.676342662632375e-05,
      "loss": 0.1593,
      "step": 4917
    },
    {
      "epoch": 3.72,
      "learning_rate": 2.675869894099849e-05,
      "loss": 0.175,
      "step": 4918
    },
    {
      "epoch": 3.72,
      "learning_rate": 2.675397125567322e-05,
      "loss": 0.1353,
      "step": 4919
    },
    {
      "epoch": 3.72,
      "learning_rate": 2.674924357034796e-05,
      "loss": 0.144,
      "step": 4920
    },
    {
      "epoch": 3.72,
      "learning_rate": 2.6744515885022692e-05,
      "loss": 0.1378,
      "step": 4921
    },
    {
      "epoch": 3.72,
      "learning_rate": 2.673978819969743e-05,
      "loss": 0.1435,
      "step": 4922
    },
    {
      "epoch": 3.72,
      "learning_rate": 2.6735060514372163e-05,
      "loss": 0.1431,
      "step": 4923
    },
    {
      "epoch": 3.72,
      "learning_rate": 2.67303328290469e-05,
      "loss": 0.1278,
      "step": 4924
    },
    {
      "epoch": 3.72,
      "learning_rate": 2.6725605143721634e-05,
      "loss": 0.1466,
      "step": 4925
    },
    {
      "epoch": 3.73,
      "learning_rate": 2.672087745839637e-05,
      "loss": 0.1592,
      "step": 4926
    },
    {
      "epoch": 3.73,
      "learning_rate": 2.672087745839637e-05,
      "loss": 0.3025,
      "step": 4927
    },
    {
      "epoch": 3.73,
      "learning_rate": 2.67161497730711e-05,
      "loss": 0.1809,
      "step": 4928
    },
    {
      "epoch": 3.73,
      "learning_rate": 2.671142208774584e-05,
      "loss": 0.163,
      "step": 4929
    },
    {
      "epoch": 3.73,
      "learning_rate": 2.6706694402420572e-05,
      "loss": 0.1812,
      "step": 4930
    },
    {
      "epoch": 3.73,
      "learning_rate": 2.670196671709531e-05,
      "loss": 0.1625,
      "step": 4931
    },
    {
      "epoch": 3.73,
      "learning_rate": 2.669723903177005e-05,
      "loss": 0.1502,
      "step": 4932
    },
    {
      "epoch": 3.73,
      "learning_rate": 2.6692511346444782e-05,
      "loss": 0.1584,
      "step": 4933
    },
    {
      "epoch": 3.73,
      "learning_rate": 2.668778366111952e-05,
      "loss": 0.145,
      "step": 4934
    },
    {
      "epoch": 3.73,
      "learning_rate": 2.6683055975794253e-05,
      "loss": 0.154,
      "step": 4935
    },
    {
      "epoch": 3.73,
      "learning_rate": 2.667832829046899e-05,
      "loss": 0.1426,
      "step": 4936
    },
    {
      "epoch": 3.73,
      "learning_rate": 2.6673600605143724e-05,
      "loss": 0.1348,
      "step": 4937
    },
    {
      "epoch": 3.73,
      "learning_rate": 2.666887291981846e-05,
      "loss": 0.161,
      "step": 4938
    },
    {
      "epoch": 3.74,
      "learning_rate": 2.666414523449319e-05,
      "loss": 0.1611,
      "step": 4939
    },
    {
      "epoch": 3.74,
      "learning_rate": 2.665941754916793e-05,
      "loss": 0.1865,
      "step": 4940
    },
    {
      "epoch": 3.74,
      "learning_rate": 2.6654689863842662e-05,
      "loss": 0.1241,
      "step": 4941
    },
    {
      "epoch": 3.74,
      "learning_rate": 2.66499621785174e-05,
      "loss": 0.1349,
      "step": 4942
    },
    {
      "epoch": 3.74,
      "learning_rate": 2.6645234493192133e-05,
      "loss": 0.1459,
      "step": 4943
    },
    {
      "epoch": 3.74,
      "learning_rate": 2.6640506807866872e-05,
      "loss": 0.1738,
      "step": 4944
    },
    {
      "epoch": 3.74,
      "learning_rate": 2.6635779122541604e-05,
      "loss": 0.148,
      "step": 4945
    },
    {
      "epoch": 3.74,
      "learning_rate": 2.6631051437216343e-05,
      "loss": 0.132,
      "step": 4946
    },
    {
      "epoch": 3.74,
      "learning_rate": 2.6626323751891075e-05,
      "loss": 0.1549,
      "step": 4947
    },
    {
      "epoch": 3.74,
      "learning_rate": 2.6621596066565814e-05,
      "loss": 0.1453,
      "step": 4948
    },
    {
      "epoch": 3.74,
      "learning_rate": 2.6616868381240546e-05,
      "loss": 0.1461,
      "step": 4949
    },
    {
      "epoch": 3.74,
      "learning_rate": 2.661214069591528e-05,
      "loss": 0.1373,
      "step": 4950
    },
    {
      "epoch": 3.74,
      "learning_rate": 2.6607413010590017e-05,
      "loss": 0.1301,
      "step": 4951
    },
    {
      "epoch": 3.75,
      "learning_rate": 2.6602685325264752e-05,
      "loss": 0.1422,
      "step": 4952
    },
    {
      "epoch": 3.75,
      "learning_rate": 2.6597957639939484e-05,
      "loss": 0.1459,
      "step": 4953
    },
    {
      "epoch": 3.75,
      "learning_rate": 2.6593229954614223e-05,
      "loss": 0.1466,
      "step": 4954
    },
    {
      "epoch": 3.75,
      "learning_rate": 2.6588502269288955e-05,
      "loss": 0.1839,
      "step": 4955
    },
    {
      "epoch": 3.75,
      "learning_rate": 2.6583774583963694e-05,
      "loss": 0.1567,
      "step": 4956
    },
    {
      "epoch": 3.75,
      "learning_rate": 2.6579046898638426e-05,
      "loss": 0.1525,
      "step": 4957
    },
    {
      "epoch": 3.75,
      "learning_rate": 2.6574319213313165e-05,
      "loss": 0.147,
      "step": 4958
    },
    {
      "epoch": 3.75,
      "learning_rate": 2.6569591527987897e-05,
      "loss": 0.1495,
      "step": 4959
    },
    {
      "epoch": 3.75,
      "learning_rate": 2.6564863842662636e-05,
      "loss": 0.168,
      "step": 4960
    },
    {
      "epoch": 3.75,
      "learning_rate": 2.6560136157337368e-05,
      "loss": 0.1413,
      "step": 4961
    },
    {
      "epoch": 3.75,
      "learning_rate": 2.6555408472012107e-05,
      "loss": 0.1574,
      "step": 4962
    },
    {
      "epoch": 3.75,
      "learning_rate": 2.655068078668684e-05,
      "loss": 0.1615,
      "step": 4963
    },
    {
      "epoch": 3.75,
      "learning_rate": 2.6545953101361574e-05,
      "loss": 0.1377,
      "step": 4964
    },
    {
      "epoch": 3.75,
      "learning_rate": 2.654122541603631e-05,
      "loss": 0.1495,
      "step": 4965
    },
    {
      "epoch": 3.76,
      "learning_rate": 2.6536497730711045e-05,
      "loss": 0.1531,
      "step": 4966
    },
    {
      "epoch": 3.76,
      "learning_rate": 2.6531770045385777e-05,
      "loss": 0.145,
      "step": 4967
    },
    {
      "epoch": 3.76,
      "learning_rate": 2.6527042360060516e-05,
      "loss": 0.14,
      "step": 4968
    },
    {
      "epoch": 3.76,
      "learning_rate": 2.6522314674735248e-05,
      "loss": 0.1541,
      "step": 4969
    },
    {
      "epoch": 3.76,
      "learning_rate": 2.6517586989409987e-05,
      "loss": 0.1503,
      "step": 4970
    },
    {
      "epoch": 3.76,
      "learning_rate": 2.651285930408472e-05,
      "loss": 0.1337,
      "step": 4971
    },
    {
      "epoch": 3.76,
      "learning_rate": 2.6508131618759458e-05,
      "loss": 0.1428,
      "step": 4972
    },
    {
      "epoch": 3.76,
      "learning_rate": 2.650340393343419e-05,
      "loss": 0.151,
      "step": 4973
    },
    {
      "epoch": 3.76,
      "learning_rate": 2.649867624810893e-05,
      "loss": 0.1368,
      "step": 4974
    },
    {
      "epoch": 3.76,
      "learning_rate": 2.649394856278366e-05,
      "loss": 0.1367,
      "step": 4975
    },
    {
      "epoch": 3.76,
      "learning_rate": 2.64892208774584e-05,
      "loss": 0.1841,
      "step": 4976
    },
    {
      "epoch": 3.76,
      "learning_rate": 2.6484493192133132e-05,
      "loss": 0.1432,
      "step": 4977
    },
    {
      "epoch": 3.76,
      "learning_rate": 2.6479765506807867e-05,
      "loss": 0.1686,
      "step": 4978
    },
    {
      "epoch": 3.77,
      "learning_rate": 2.6475037821482603e-05,
      "loss": 0.1416,
      "step": 4979
    },
    {
      "epoch": 3.77,
      "learning_rate": 2.6470310136157338e-05,
      "loss": 0.1578,
      "step": 4980
    },
    {
      "epoch": 3.77,
      "learning_rate": 2.646558245083207e-05,
      "loss": 0.1626,
      "step": 4981
    },
    {
      "epoch": 3.77,
      "learning_rate": 2.646085476550681e-05,
      "loss": 0.1557,
      "step": 4982
    },
    {
      "epoch": 3.77,
      "learning_rate": 2.645612708018154e-05,
      "loss": 0.1491,
      "step": 4983
    },
    {
      "epoch": 3.77,
      "learning_rate": 2.645139939485628e-05,
      "loss": 0.1667,
      "step": 4984
    },
    {
      "epoch": 3.77,
      "learning_rate": 2.6446671709531012e-05,
      "loss": 0.1578,
      "step": 4985
    },
    {
      "epoch": 3.77,
      "learning_rate": 2.644194402420575e-05,
      "loss": 0.1532,
      "step": 4986
    },
    {
      "epoch": 3.77,
      "learning_rate": 2.6437216338880483e-05,
      "loss": 0.1406,
      "step": 4987
    },
    {
      "epoch": 3.77,
      "learning_rate": 2.643248865355522e-05,
      "loss": 0.1655,
      "step": 4988
    },
    {
      "epoch": 3.77,
      "learning_rate": 2.6427760968229954e-05,
      "loss": 0.1692,
      "step": 4989
    },
    {
      "epoch": 3.77,
      "learning_rate": 2.6423033282904693e-05,
      "loss": 0.171,
      "step": 4990
    },
    {
      "epoch": 3.77,
      "learning_rate": 2.6418305597579428e-05,
      "loss": 0.1708,
      "step": 4991
    },
    {
      "epoch": 3.78,
      "learning_rate": 2.641357791225416e-05,
      "loss": 0.1521,
      "step": 4992
    },
    {
      "epoch": 3.78,
      "learning_rate": 2.64088502269289e-05,
      "loss": 0.1472,
      "step": 4993
    },
    {
      "epoch": 3.78,
      "learning_rate": 2.640412254160363e-05,
      "loss": 0.1319,
      "step": 4994
    },
    {
      "epoch": 3.78,
      "learning_rate": 2.639939485627837e-05,
      "loss": 0.1369,
      "step": 4995
    },
    {
      "epoch": 3.78,
      "learning_rate": 2.6394667170953102e-05,
      "loss": 0.1599,
      "step": 4996
    },
    {
      "epoch": 3.78,
      "learning_rate": 2.638993948562784e-05,
      "loss": 0.1736,
      "step": 4997
    },
    {
      "epoch": 3.78,
      "learning_rate": 2.6385211800302573e-05,
      "loss": 0.1408,
      "step": 4998
    },
    {
      "epoch": 3.78,
      "learning_rate": 2.638048411497731e-05,
      "loss": 0.1312,
      "step": 4999
    },
    {
      "epoch": 3.78,
      "learning_rate": 2.6375756429652044e-05,
      "loss": 0.1483,
      "step": 5000
    },
    {
      "epoch": 3.78,
      "learning_rate": 2.6371028744326782e-05,
      "loss": 0.1633,
      "step": 5001
    },
    {
      "epoch": 3.78,
      "learning_rate": 2.6366301059001515e-05,
      "loss": 0.1408,
      "step": 5002
    },
    {
      "epoch": 3.78,
      "learning_rate": 2.636157337367625e-05,
      "loss": 0.1352,
      "step": 5003
    },
    {
      "epoch": 3.78,
      "learning_rate": 2.6356845688350985e-05,
      "loss": 0.1189,
      "step": 5004
    },
    {
      "epoch": 3.79,
      "learning_rate": 2.635211800302572e-05,
      "loss": 0.1509,
      "step": 5005
    },
    {
      "epoch": 3.79,
      "learning_rate": 2.6347390317700453e-05,
      "loss": 0.1578,
      "step": 5006
    },
    {
      "epoch": 3.79,
      "learning_rate": 2.6342662632375192e-05,
      "loss": 0.1538,
      "step": 5007
    },
    {
      "epoch": 3.79,
      "learning_rate": 2.6337934947049924e-05,
      "loss": 0.141,
      "step": 5008
    },
    {
      "epoch": 3.79,
      "learning_rate": 2.6333207261724663e-05,
      "loss": 0.1599,
      "step": 5009
    },
    {
      "epoch": 3.79,
      "learning_rate": 2.6328479576399395e-05,
      "loss": 0.131,
      "step": 5010
    },
    {
      "epoch": 3.79,
      "learning_rate": 2.6323751891074134e-05,
      "loss": 0.1545,
      "step": 5011
    },
    {
      "epoch": 3.79,
      "learning_rate": 2.6319024205748866e-05,
      "loss": 0.1452,
      "step": 5012
    },
    {
      "epoch": 3.79,
      "learning_rate": 2.6314296520423604e-05,
      "loss": 0.1557,
      "step": 5013
    },
    {
      "epoch": 3.79,
      "learning_rate": 2.6309568835098336e-05,
      "loss": 0.147,
      "step": 5014
    },
    {
      "epoch": 3.79,
      "learning_rate": 2.6304841149773075e-05,
      "loss": 0.1561,
      "step": 5015
    },
    {
      "epoch": 3.79,
      "learning_rate": 2.6300113464447807e-05,
      "loss": 0.148,
      "step": 5016
    },
    {
      "epoch": 3.79,
      "learning_rate": 2.6295385779122543e-05,
      "loss": 0.145,
      "step": 5017
    },
    {
      "epoch": 3.8,
      "learning_rate": 2.6290658093797278e-05,
      "loss": 0.1668,
      "step": 5018
    },
    {
      "epoch": 3.8,
      "learning_rate": 2.6285930408472014e-05,
      "loss": 0.1383,
      "step": 5019
    },
    {
      "epoch": 3.8,
      "learning_rate": 2.6281202723146746e-05,
      "loss": 0.1489,
      "step": 5020
    },
    {
      "epoch": 3.8,
      "learning_rate": 2.6276475037821485e-05,
      "loss": 0.1731,
      "step": 5021
    },
    {
      "epoch": 3.8,
      "learning_rate": 2.6271747352496217e-05,
      "loss": 0.1433,
      "step": 5022
    },
    {
      "epoch": 3.8,
      "learning_rate": 2.6267019667170956e-05,
      "loss": 0.136,
      "step": 5023
    },
    {
      "epoch": 3.8,
      "learning_rate": 2.6262291981845688e-05,
      "loss": 0.1366,
      "step": 5024
    },
    {
      "epoch": 3.8,
      "learning_rate": 2.6257564296520426e-05,
      "loss": 0.1445,
      "step": 5025
    },
    {
      "epoch": 3.8,
      "learning_rate": 2.625283661119516e-05,
      "loss": 0.1385,
      "step": 5026
    },
    {
      "epoch": 3.8,
      "learning_rate": 2.6248108925869897e-05,
      "loss": 0.1769,
      "step": 5027
    },
    {
      "epoch": 3.8,
      "learning_rate": 2.624338124054463e-05,
      "loss": 0.1384,
      "step": 5028
    },
    {
      "epoch": 3.8,
      "learning_rate": 2.6238653555219368e-05,
      "loss": 0.1372,
      "step": 5029
    },
    {
      "epoch": 3.8,
      "learning_rate": 2.62339258698941e-05,
      "loss": 0.1356,
      "step": 5030
    },
    {
      "epoch": 3.8,
      "learning_rate": 2.6229198184568836e-05,
      "loss": 0.1285,
      "step": 5031
    },
    {
      "epoch": 3.81,
      "learning_rate": 2.622447049924357e-05,
      "loss": 0.1554,
      "step": 5032
    },
    {
      "epoch": 3.81,
      "learning_rate": 2.6219742813918307e-05,
      "loss": 0.1523,
      "step": 5033
    },
    {
      "epoch": 3.81,
      "learning_rate": 2.621501512859304e-05,
      "loss": 0.1331,
      "step": 5034
    },
    {
      "epoch": 3.81,
      "learning_rate": 2.6210287443267777e-05,
      "loss": 0.1851,
      "step": 5035
    },
    {
      "epoch": 3.81,
      "learning_rate": 2.620555975794251e-05,
      "loss": 0.1326,
      "step": 5036
    },
    {
      "epoch": 3.81,
      "learning_rate": 2.620083207261725e-05,
      "loss": 0.1715,
      "step": 5037
    },
    {
      "epoch": 3.81,
      "learning_rate": 2.619610438729198e-05,
      "loss": 0.1515,
      "step": 5038
    },
    {
      "epoch": 3.81,
      "learning_rate": 2.619137670196672e-05,
      "loss": 0.1339,
      "step": 5039
    },
    {
      "epoch": 3.81,
      "learning_rate": 2.618664901664145e-05,
      "loss": 0.1433,
      "step": 5040
    },
    {
      "epoch": 3.81,
      "learning_rate": 2.618192133131619e-05,
      "loss": 0.1603,
      "step": 5041
    },
    {
      "epoch": 3.81,
      "learning_rate": 2.6177193645990922e-05,
      "loss": 0.142,
      "step": 5042
    },
    {
      "epoch": 3.81,
      "learning_rate": 2.617246596066566e-05,
      "loss": 0.1488,
      "step": 5043
    },
    {
      "epoch": 3.81,
      "learning_rate": 2.6167738275340393e-05,
      "loss": 0.1606,
      "step": 5044
    },
    {
      "epoch": 3.82,
      "learning_rate": 2.616301059001513e-05,
      "loss": 0.1502,
      "step": 5045
    },
    {
      "epoch": 3.82,
      "learning_rate": 2.6158282904689864e-05,
      "loss": 0.153,
      "step": 5046
    },
    {
      "epoch": 3.82,
      "learning_rate": 2.61535552193646e-05,
      "loss": 0.1403,
      "step": 5047
    },
    {
      "epoch": 3.82,
      "learning_rate": 2.614882753403933e-05,
      "loss": 0.1521,
      "step": 5048
    },
    {
      "epoch": 3.82,
      "learning_rate": 2.614409984871407e-05,
      "loss": 0.1393,
      "step": 5049
    },
    {
      "epoch": 3.82,
      "learning_rate": 2.6139372163388802e-05,
      "loss": 0.1445,
      "step": 5050
    },
    {
      "epoch": 3.82,
      "learning_rate": 2.613464447806354e-05,
      "loss": 0.1542,
      "step": 5051
    },
    {
      "epoch": 3.82,
      "learning_rate": 2.612991679273828e-05,
      "loss": 0.1352,
      "step": 5052
    },
    {
      "epoch": 3.82,
      "learning_rate": 2.6125189107413012e-05,
      "loss": 0.125,
      "step": 5053
    },
    {
      "epoch": 3.82,
      "learning_rate": 2.612046142208775e-05,
      "loss": 0.156,
      "step": 5054
    },
    {
      "epoch": 3.82,
      "learning_rate": 2.6115733736762483e-05,
      "loss": 0.1469,
      "step": 5055
    },
    {
      "epoch": 3.82,
      "learning_rate": 2.611100605143722e-05,
      "loss": 0.1389,
      "step": 5056
    },
    {
      "epoch": 3.82,
      "learning_rate": 2.6106278366111954e-05,
      "loss": 0.1731,
      "step": 5057
    },
    {
      "epoch": 3.83,
      "learning_rate": 2.610155068078669e-05,
      "loss": 0.1454,
      "step": 5058
    },
    {
      "epoch": 3.83,
      "learning_rate": 2.609682299546142e-05,
      "loss": 0.131,
      "step": 5059
    },
    {
      "epoch": 3.83,
      "learning_rate": 2.609209531013616e-05,
      "loss": 0.1281,
      "step": 5060
    },
    {
      "epoch": 3.83,
      "learning_rate": 2.6087367624810892e-05,
      "loss": 0.1671,
      "step": 5061
    },
    {
      "epoch": 3.83,
      "learning_rate": 2.608263993948563e-05,
      "loss": 0.1489,
      "step": 5062
    },
    {
      "epoch": 3.83,
      "learning_rate": 2.6077912254160363e-05,
      "loss": 0.1573,
      "step": 5063
    },
    {
      "epoch": 3.83,
      "learning_rate": 2.6073184568835102e-05,
      "loss": 0.1552,
      "step": 5064
    },
    {
      "epoch": 3.83,
      "learning_rate": 2.6068456883509834e-05,
      "loss": 0.1641,
      "step": 5065
    },
    {
      "epoch": 3.83,
      "learning_rate": 2.6063729198184573e-05,
      "loss": 0.1517,
      "step": 5066
    },
    {
      "epoch": 3.83,
      "learning_rate": 2.6059001512859305e-05,
      "loss": 0.1458,
      "step": 5067
    },
    {
      "epoch": 3.83,
      "learning_rate": 2.6054273827534044e-05,
      "loss": 0.1873,
      "step": 5068
    },
    {
      "epoch": 3.83,
      "learning_rate": 2.6049546142208776e-05,
      "loss": 0.1639,
      "step": 5069
    },
    {
      "epoch": 3.83,
      "learning_rate": 2.604481845688351e-05,
      "loss": 0.1424,
      "step": 5070
    },
    {
      "epoch": 3.84,
      "learning_rate": 2.6040090771558247e-05,
      "loss": 0.1631,
      "step": 5071
    },
    {
      "epoch": 3.84,
      "learning_rate": 2.6035363086232982e-05,
      "loss": 0.1463,
      "step": 5072
    },
    {
      "epoch": 3.84,
      "learning_rate": 2.6030635400907714e-05,
      "loss": 0.1338,
      "step": 5073
    },
    {
      "epoch": 3.84,
      "learning_rate": 2.6025907715582453e-05,
      "loss": 0.1565,
      "step": 5074
    },
    {
      "epoch": 3.84,
      "learning_rate": 2.6021180030257185e-05,
      "loss": 0.1505,
      "step": 5075
    },
    {
      "epoch": 3.84,
      "learning_rate": 2.6016452344931924e-05,
      "loss": 0.1713,
      "step": 5076
    },
    {
      "epoch": 3.84,
      "learning_rate": 2.6011724659606656e-05,
      "loss": 0.1343,
      "step": 5077
    },
    {
      "epoch": 3.84,
      "learning_rate": 2.6006996974281395e-05,
      "loss": 0.1326,
      "step": 5078
    },
    {
      "epoch": 3.84,
      "learning_rate": 2.6002269288956127e-05,
      "loss": 0.1556,
      "step": 5079
    },
    {
      "epoch": 3.84,
      "learning_rate": 2.5997541603630866e-05,
      "loss": 0.1432,
      "step": 5080
    },
    {
      "epoch": 3.84,
      "learning_rate": 2.5992813918305598e-05,
      "loss": 0.1444,
      "step": 5081
    },
    {
      "epoch": 3.84,
      "learning_rate": 2.5988086232980337e-05,
      "loss": 0.1235,
      "step": 5082
    },
    {
      "epoch": 3.84,
      "learning_rate": 2.598335854765507e-05,
      "loss": 0.1463,
      "step": 5083
    },
    {
      "epoch": 3.84,
      "learning_rate": 2.5978630862329804e-05,
      "loss": 0.1615,
      "step": 5084
    },
    {
      "epoch": 3.85,
      "learning_rate": 2.597390317700454e-05,
      "loss": 0.159,
      "step": 5085
    },
    {
      "epoch": 3.85,
      "learning_rate": 2.5969175491679275e-05,
      "loss": 0.1551,
      "step": 5086
    },
    {
      "epoch": 3.85,
      "learning_rate": 2.5964447806354007e-05,
      "loss": 0.155,
      "step": 5087
    },
    {
      "epoch": 3.85,
      "learning_rate": 2.5959720121028746e-05,
      "loss": 0.135,
      "step": 5088
    },
    {
      "epoch": 3.85,
      "learning_rate": 2.5954992435703478e-05,
      "loss": 0.1648,
      "step": 5089
    },
    {
      "epoch": 3.85,
      "learning_rate": 2.5950264750378217e-05,
      "loss": 0.1681,
      "step": 5090
    },
    {
      "epoch": 3.85,
      "learning_rate": 2.594553706505295e-05,
      "loss": 0.1527,
      "step": 5091
    },
    {
      "epoch": 3.85,
      "learning_rate": 2.5940809379727688e-05,
      "loss": 0.1638,
      "step": 5092
    },
    {
      "epoch": 3.85,
      "learning_rate": 2.593608169440242e-05,
      "loss": 0.136,
      "step": 5093
    },
    {
      "epoch": 3.85,
      "learning_rate": 2.593135400907716e-05,
      "loss": 0.1563,
      "step": 5094
    },
    {
      "epoch": 3.85,
      "learning_rate": 2.592662632375189e-05,
      "loss": 0.1534,
      "step": 5095
    },
    {
      "epoch": 3.85,
      "learning_rate": 2.592189863842663e-05,
      "loss": 0.1479,
      "step": 5096
    },
    {
      "epoch": 3.85,
      "learning_rate": 2.591717095310136e-05,
      "loss": 0.1729,
      "step": 5097
    },
    {
      "epoch": 3.86,
      "learning_rate": 2.5912443267776097e-05,
      "loss": 0.1423,
      "step": 5098
    },
    {
      "epoch": 3.86,
      "learning_rate": 2.5907715582450833e-05,
      "loss": 0.1467,
      "step": 5099
    },
    {
      "epoch": 3.86,
      "learning_rate": 2.5902987897125568e-05,
      "loss": 0.1577,
      "step": 5100
    },
    {
      "epoch": 3.86,
      "learning_rate": 2.58982602118003e-05,
      "loss": 0.1592,
      "step": 5101
    },
    {
      "epoch": 3.86,
      "learning_rate": 2.589353252647504e-05,
      "loss": 0.1456,
      "step": 5102
    },
    {
      "epoch": 3.86,
      "learning_rate": 2.588880484114977e-05,
      "loss": 0.1507,
      "step": 5103
    },
    {
      "epoch": 3.86,
      "learning_rate": 2.588407715582451e-05,
      "loss": 0.1413,
      "step": 5104
    },
    {
      "epoch": 3.86,
      "learning_rate": 2.5879349470499242e-05,
      "loss": 0.1237,
      "step": 5105
    },
    {
      "epoch": 3.86,
      "learning_rate": 2.587462178517398e-05,
      "loss": 0.1335,
      "step": 5106
    },
    {
      "epoch": 3.86,
      "learning_rate": 2.5869894099848713e-05,
      "loss": 0.1253,
      "step": 5107
    },
    {
      "epoch": 3.86,
      "learning_rate": 2.586516641452345e-05,
      "loss": 0.1573,
      "step": 5108
    },
    {
      "epoch": 3.86,
      "learning_rate": 2.5860438729198184e-05,
      "loss": 0.1376,
      "step": 5109
    },
    {
      "epoch": 3.86,
      "learning_rate": 2.5855711043872922e-05,
      "loss": 0.1372,
      "step": 5110
    },
    {
      "epoch": 3.87,
      "learning_rate": 2.5850983358547658e-05,
      "loss": 0.1637,
      "step": 5111
    },
    {
      "epoch": 3.87,
      "learning_rate": 2.584625567322239e-05,
      "loss": 0.1324,
      "step": 5112
    },
    {
      "epoch": 3.87,
      "learning_rate": 2.584152798789713e-05,
      "loss": 0.152,
      "step": 5113
    },
    {
      "epoch": 3.87,
      "learning_rate": 2.583680030257186e-05,
      "loss": 0.1829,
      "step": 5114
    },
    {
      "epoch": 3.87,
      "learning_rate": 2.58320726172466e-05,
      "loss": 0.1242,
      "step": 5115
    },
    {
      "epoch": 3.87,
      "learning_rate": 2.5827344931921332e-05,
      "loss": 0.1475,
      "step": 5116
    },
    {
      "epoch": 3.87,
      "learning_rate": 2.582261724659607e-05,
      "loss": 0.1277,
      "step": 5117
    },
    {
      "epoch": 3.87,
      "learning_rate": 2.5817889561270803e-05,
      "loss": 0.1475,
      "step": 5118
    },
    {
      "epoch": 3.87,
      "learning_rate": 2.581316187594554e-05,
      "loss": 0.1366,
      "step": 5119
    },
    {
      "epoch": 3.87,
      "learning_rate": 2.5808434190620274e-05,
      "loss": 0.135,
      "step": 5120
    },
    {
      "epoch": 3.87,
      "learning_rate": 2.5803706505295012e-05,
      "loss": 0.1718,
      "step": 5121
    },
    {
      "epoch": 3.87,
      "learning_rate": 2.5798978819969744e-05,
      "loss": 0.1379,
      "step": 5122
    },
    {
      "epoch": 3.87,
      "learning_rate": 2.579425113464448e-05,
      "loss": 0.1682,
      "step": 5123
    },
    {
      "epoch": 3.88,
      "learning_rate": 2.5789523449319215e-05,
      "loss": 0.1493,
      "step": 5124
    },
    {
      "epoch": 3.88,
      "learning_rate": 2.578479576399395e-05,
      "loss": 0.1345,
      "step": 5125
    },
    {
      "epoch": 3.88,
      "learning_rate": 2.5780068078668683e-05,
      "loss": 0.1724,
      "step": 5126
    },
    {
      "epoch": 3.88,
      "learning_rate": 2.5775340393343422e-05,
      "loss": 0.1399,
      "step": 5127
    },
    {
      "epoch": 3.88,
      "learning_rate": 2.5770612708018154e-05,
      "loss": 0.1733,
      "step": 5128
    },
    {
      "epoch": 3.88,
      "learning_rate": 2.5765885022692893e-05,
      "loss": 0.1725,
      "step": 5129
    },
    {
      "epoch": 3.88,
      "learning_rate": 2.5761157337367625e-05,
      "loss": 0.1392,
      "step": 5130
    },
    {
      "epoch": 3.88,
      "learning_rate": 2.5756429652042363e-05,
      "loss": 0.1565,
      "step": 5131
    },
    {
      "epoch": 3.88,
      "learning_rate": 2.5751701966717096e-05,
      "loss": 0.151,
      "step": 5132
    },
    {
      "epoch": 3.88,
      "learning_rate": 2.5746974281391834e-05,
      "loss": 0.1485,
      "step": 5133
    },
    {
      "epoch": 3.88,
      "learning_rate": 2.5742246596066566e-05,
      "loss": 0.1421,
      "step": 5134
    },
    {
      "epoch": 3.88,
      "learning_rate": 2.5737518910741305e-05,
      "loss": 0.1469,
      "step": 5135
    },
    {
      "epoch": 3.88,
      "learning_rate": 2.5732791225416037e-05,
      "loss": 0.1278,
      "step": 5136
    },
    {
      "epoch": 3.89,
      "learning_rate": 2.5728063540090773e-05,
      "loss": 0.134,
      "step": 5137
    },
    {
      "epoch": 3.89,
      "learning_rate": 2.5723335854765508e-05,
      "loss": 0.155,
      "step": 5138
    },
    {
      "epoch": 3.89,
      "learning_rate": 2.5718608169440244e-05,
      "loss": 0.1532,
      "step": 5139
    },
    {
      "epoch": 3.89,
      "learning_rate": 2.5713880484114976e-05,
      "loss": 0.1542,
      "step": 5140
    },
    {
      "epoch": 3.89,
      "learning_rate": 2.5709152798789715e-05,
      "loss": 0.1422,
      "step": 5141
    },
    {
      "epoch": 3.89,
      "learning_rate": 2.5704425113464447e-05,
      "loss": 0.1362,
      "step": 5142
    },
    {
      "epoch": 3.89,
      "learning_rate": 2.5699697428139185e-05,
      "loss": 0.1312,
      "step": 5143
    },
    {
      "epoch": 3.89,
      "learning_rate": 2.5694969742813918e-05,
      "loss": 0.1531,
      "step": 5144
    },
    {
      "epoch": 3.89,
      "learning_rate": 2.5690242057488656e-05,
      "loss": 0.1347,
      "step": 5145
    },
    {
      "epoch": 3.89,
      "learning_rate": 2.568551437216339e-05,
      "loss": 0.1572,
      "step": 5146
    },
    {
      "epoch": 3.89,
      "learning_rate": 2.5680786686838127e-05,
      "loss": 0.1619,
      "step": 5147
    },
    {
      "epoch": 3.89,
      "learning_rate": 2.567605900151286e-05,
      "loss": 0.1435,
      "step": 5148
    },
    {
      "epoch": 3.89,
      "learning_rate": 2.5671331316187598e-05,
      "loss": 0.1168,
      "step": 5149
    },
    {
      "epoch": 3.89,
      "learning_rate": 2.566660363086233e-05,
      "loss": 0.1555,
      "step": 5150
    },
    {
      "epoch": 3.9,
      "learning_rate": 2.5661875945537066e-05,
      "loss": 0.1318,
      "step": 5151
    },
    {
      "epoch": 3.9,
      "learning_rate": 2.56571482602118e-05,
      "loss": 0.163,
      "step": 5152
    },
    {
      "epoch": 3.9,
      "learning_rate": 2.5652420574886537e-05,
      "loss": 0.1446,
      "step": 5153
    },
    {
      "epoch": 3.9,
      "learning_rate": 2.564769288956127e-05,
      "loss": 0.1531,
      "step": 5154
    },
    {
      "epoch": 3.9,
      "learning_rate": 2.5642965204236007e-05,
      "loss": 0.1334,
      "step": 5155
    },
    {
      "epoch": 3.9,
      "learning_rate": 2.563823751891074e-05,
      "loss": 0.1384,
      "step": 5156
    },
    {
      "epoch": 3.9,
      "learning_rate": 2.563350983358548e-05,
      "loss": 0.1364,
      "step": 5157
    },
    {
      "epoch": 3.9,
      "learning_rate": 2.562878214826021e-05,
      "loss": 0.1585,
      "step": 5158
    },
    {
      "epoch": 3.9,
      "learning_rate": 2.562405446293495e-05,
      "loss": 0.1344,
      "step": 5159
    },
    {
      "epoch": 3.9,
      "learning_rate": 2.561932677760968e-05,
      "loss": 0.1317,
      "step": 5160
    },
    {
      "epoch": 3.9,
      "learning_rate": 2.561459909228442e-05,
      "loss": 0.1338,
      "step": 5161
    },
    {
      "epoch": 3.9,
      "learning_rate": 2.5609871406959152e-05,
      "loss": 0.1511,
      "step": 5162
    },
    {
      "epoch": 3.9,
      "learning_rate": 2.560514372163389e-05,
      "loss": 0.1313,
      "step": 5163
    },
    {
      "epoch": 3.91,
      "learning_rate": 2.5600416036308623e-05,
      "loss": 0.1552,
      "step": 5164
    },
    {
      "epoch": 3.91,
      "learning_rate": 2.559568835098336e-05,
      "loss": 0.1556,
      "step": 5165
    },
    {
      "epoch": 3.91,
      "learning_rate": 2.5590960665658094e-05,
      "loss": 0.1504,
      "step": 5166
    },
    {
      "epoch": 3.91,
      "learning_rate": 2.558623298033283e-05,
      "loss": 0.1287,
      "step": 5167
    },
    {
      "epoch": 3.91,
      "learning_rate": 2.558150529500756e-05,
      "loss": 0.1306,
      "step": 5168
    },
    {
      "epoch": 3.91,
      "learning_rate": 2.55767776096823e-05,
      "loss": 0.1292,
      "step": 5169
    },
    {
      "epoch": 3.91,
      "learning_rate": 2.5572049924357032e-05,
      "loss": 0.1398,
      "step": 5170
    },
    {
      "epoch": 3.91,
      "learning_rate": 2.556732223903177e-05,
      "loss": 0.1641,
      "step": 5171
    },
    {
      "epoch": 3.91,
      "learning_rate": 2.556259455370651e-05,
      "loss": 0.1615,
      "step": 5172
    },
    {
      "epoch": 3.91,
      "learning_rate": 2.5557866868381242e-05,
      "loss": 0.1381,
      "step": 5173
    },
    {
      "epoch": 3.91,
      "learning_rate": 2.555313918305598e-05,
      "loss": 0.1219,
      "step": 5174
    },
    {
      "epoch": 3.91,
      "learning_rate": 2.5548411497730713e-05,
      "loss": 0.1437,
      "step": 5175
    },
    {
      "epoch": 3.91,
      "learning_rate": 2.554368381240545e-05,
      "loss": 0.1528,
      "step": 5176
    },
    {
      "epoch": 3.92,
      "learning_rate": 2.5538956127080184e-05,
      "loss": 0.1438,
      "step": 5177
    },
    {
      "epoch": 3.92,
      "learning_rate": 2.553422844175492e-05,
      "loss": 0.128,
      "step": 5178
    },
    {
      "epoch": 3.92,
      "learning_rate": 2.552950075642965e-05,
      "loss": 0.1455,
      "step": 5179
    },
    {
      "epoch": 3.92,
      "learning_rate": 2.552477307110439e-05,
      "loss": 0.1526,
      "step": 5180
    },
    {
      "epoch": 3.92,
      "learning_rate": 2.5520045385779122e-05,
      "loss": 0.1386,
      "step": 5181
    },
    {
      "epoch": 3.92,
      "learning_rate": 2.551531770045386e-05,
      "loss": 0.1704,
      "step": 5182
    },
    {
      "epoch": 3.92,
      "learning_rate": 2.5510590015128593e-05,
      "loss": 0.1526,
      "step": 5183
    },
    {
      "epoch": 3.92,
      "learning_rate": 2.5505862329803332e-05,
      "loss": 0.1551,
      "step": 5184
    },
    {
      "epoch": 3.92,
      "learning_rate": 2.5501134644478064e-05,
      "loss": 0.1295,
      "step": 5185
    },
    {
      "epoch": 3.92,
      "learning_rate": 2.5496406959152803e-05,
      "loss": 0.143,
      "step": 5186
    },
    {
      "epoch": 3.92,
      "learning_rate": 2.5491679273827535e-05,
      "loss": 0.1712,
      "step": 5187
    },
    {
      "epoch": 3.92,
      "learning_rate": 2.5486951588502274e-05,
      "loss": 0.1427,
      "step": 5188
    },
    {
      "epoch": 3.92,
      "learning_rate": 2.5482223903177006e-05,
      "loss": 0.1461,
      "step": 5189
    },
    {
      "epoch": 3.93,
      "learning_rate": 2.547749621785174e-05,
      "loss": 0.1381,
      "step": 5190
    },
    {
      "epoch": 3.93,
      "learning_rate": 2.5472768532526477e-05,
      "loss": 0.1439,
      "step": 5191
    },
    {
      "epoch": 3.93,
      "learning_rate": 2.5468040847201212e-05,
      "loss": 0.1316,
      "step": 5192
    },
    {
      "epoch": 3.93,
      "learning_rate": 2.5463313161875944e-05,
      "loss": 0.164,
      "step": 5193
    },
    {
      "epoch": 3.93,
      "learning_rate": 2.5458585476550683e-05,
      "loss": 0.1377,
      "step": 5194
    },
    {
      "epoch": 3.93,
      "learning_rate": 2.5453857791225415e-05,
      "loss": 0.1395,
      "step": 5195
    },
    {
      "epoch": 3.93,
      "learning_rate": 2.5449130105900154e-05,
      "loss": 0.1533,
      "step": 5196
    },
    {
      "epoch": 3.93,
      "learning_rate": 2.5444402420574886e-05,
      "loss": 0.1466,
      "step": 5197
    },
    {
      "epoch": 3.93,
      "learning_rate": 2.5439674735249625e-05,
      "loss": 0.1444,
      "step": 5198
    },
    {
      "epoch": 3.93,
      "learning_rate": 2.5434947049924357e-05,
      "loss": 0.1504,
      "step": 5199
    },
    {
      "epoch": 3.93,
      "learning_rate": 2.5430219364599096e-05,
      "loss": 0.1732,
      "step": 5200
    },
    {
      "epoch": 3.93,
      "learning_rate": 2.5425491679273828e-05,
      "loss": 0.1485,
      "step": 5201
    },
    {
      "epoch": 3.93,
      "learning_rate": 2.5420763993948567e-05,
      "loss": 0.1768,
      "step": 5202
    },
    {
      "epoch": 3.93,
      "learning_rate": 2.54160363086233e-05,
      "loss": 0.1445,
      "step": 5203
    },
    {
      "epoch": 3.94,
      "learning_rate": 2.5411308623298034e-05,
      "loss": 0.1495,
      "step": 5204
    },
    {
      "epoch": 3.94,
      "learning_rate": 2.540658093797277e-05,
      "loss": 0.1597,
      "step": 5205
    },
    {
      "epoch": 3.94,
      "learning_rate": 2.5401853252647505e-05,
      "loss": 0.1554,
      "step": 5206
    },
    {
      "epoch": 3.94,
      "learning_rate": 2.5397125567322237e-05,
      "loss": 0.1347,
      "step": 5207
    },
    {
      "epoch": 3.94,
      "learning_rate": 2.5392397881996976e-05,
      "loss": 0.1499,
      "step": 5208
    },
    {
      "epoch": 3.94,
      "learning_rate": 2.5387670196671708e-05,
      "loss": 0.1398,
      "step": 5209
    },
    {
      "epoch": 3.94,
      "learning_rate": 2.5382942511346447e-05,
      "loss": 0.1577,
      "step": 5210
    },
    {
      "epoch": 3.94,
      "learning_rate": 2.537821482602118e-05,
      "loss": 0.1283,
      "step": 5211
    },
    {
      "epoch": 3.94,
      "learning_rate": 2.5373487140695918e-05,
      "loss": 0.1261,
      "step": 5212
    },
    {
      "epoch": 3.94,
      "learning_rate": 2.536875945537065e-05,
      "loss": 0.1549,
      "step": 5213
    },
    {
      "epoch": 3.94,
      "learning_rate": 2.536403177004539e-05,
      "loss": 0.1408,
      "step": 5214
    },
    {
      "epoch": 3.94,
      "learning_rate": 2.535930408472012e-05,
      "loss": 0.1284,
      "step": 5215
    },
    {
      "epoch": 3.94,
      "learning_rate": 2.535457639939486e-05,
      "loss": 0.134,
      "step": 5216
    },
    {
      "epoch": 3.95,
      "learning_rate": 2.534984871406959e-05,
      "loss": 0.1602,
      "step": 5217
    },
    {
      "epoch": 3.95,
      "learning_rate": 2.5345121028744327e-05,
      "loss": 0.1265,
      "step": 5218
    },
    {
      "epoch": 3.95,
      "learning_rate": 2.5340393343419063e-05,
      "loss": 0.1412,
      "step": 5219
    },
    {
      "epoch": 3.95,
      "learning_rate": 2.5335665658093798e-05,
      "loss": 0.1745,
      "step": 5220
    },
    {
      "epoch": 3.95,
      "learning_rate": 2.533093797276853e-05,
      "loss": 0.1568,
      "step": 5221
    },
    {
      "epoch": 3.95,
      "learning_rate": 2.532621028744327e-05,
      "loss": 0.1552,
      "step": 5222
    },
    {
      "epoch": 3.95,
      "learning_rate": 2.5321482602118e-05,
      "loss": 0.1413,
      "step": 5223
    },
    {
      "epoch": 3.95,
      "learning_rate": 2.531675491679274e-05,
      "loss": 0.1498,
      "step": 5224
    },
    {
      "epoch": 3.95,
      "learning_rate": 2.5312027231467472e-05,
      "loss": 0.1351,
      "step": 5225
    },
    {
      "epoch": 3.95,
      "learning_rate": 2.530729954614221e-05,
      "loss": 0.1516,
      "step": 5226
    },
    {
      "epoch": 3.95,
      "learning_rate": 2.5302571860816943e-05,
      "loss": 0.1474,
      "step": 5227
    },
    {
      "epoch": 3.95,
      "learning_rate": 2.529784417549168e-05,
      "loss": 0.1527,
      "step": 5228
    },
    {
      "epoch": 3.95,
      "learning_rate": 2.5293116490166414e-05,
      "loss": 0.1646,
      "step": 5229
    },
    {
      "epoch": 3.96,
      "learning_rate": 2.5288388804841152e-05,
      "loss": 0.1502,
      "step": 5230
    },
    {
      "epoch": 3.96,
      "learning_rate": 2.5283661119515888e-05,
      "loss": 0.1421,
      "step": 5231
    },
    {
      "epoch": 3.96,
      "learning_rate": 2.527893343419062e-05,
      "loss": 0.1644,
      "step": 5232
    },
    {
      "epoch": 3.96,
      "learning_rate": 2.527420574886536e-05,
      "loss": 0.1466,
      "step": 5233
    },
    {
      "epoch": 3.96,
      "learning_rate": 2.526947806354009e-05,
      "loss": 0.1483,
      "step": 5234
    },
    {
      "epoch": 3.96,
      "learning_rate": 2.526475037821483e-05,
      "loss": 0.152,
      "step": 5235
    },
    {
      "epoch": 3.96,
      "learning_rate": 2.5260022692889562e-05,
      "loss": 0.154,
      "step": 5236
    },
    {
      "epoch": 3.96,
      "learning_rate": 2.52552950075643e-05,
      "loss": 0.1797,
      "step": 5237
    },
    {
      "epoch": 3.96,
      "learning_rate": 2.5250567322239033e-05,
      "loss": 0.1673,
      "step": 5238
    },
    {
      "epoch": 3.96,
      "learning_rate": 2.524583963691377e-05,
      "loss": 0.1699,
      "step": 5239
    },
    {
      "epoch": 3.96,
      "learning_rate": 2.5241111951588504e-05,
      "loss": 0.141,
      "step": 5240
    },
    {
      "epoch": 3.96,
      "learning_rate": 2.5236384266263242e-05,
      "loss": 0.1272,
      "step": 5241
    },
    {
      "epoch": 3.96,
      "learning_rate": 2.5231656580937974e-05,
      "loss": 0.1502,
      "step": 5242
    },
    {
      "epoch": 3.97,
      "learning_rate": 2.522692889561271e-05,
      "loss": 0.1644,
      "step": 5243
    },
    {
      "epoch": 3.97,
      "learning_rate": 2.5222201210287445e-05,
      "loss": 0.1565,
      "step": 5244
    },
    {
      "epoch": 3.97,
      "learning_rate": 2.521747352496218e-05,
      "loss": 0.1444,
      "step": 5245
    },
    {
      "epoch": 3.97,
      "learning_rate": 2.5212745839636913e-05,
      "loss": 0.1607,
      "step": 5246
    },
    {
      "epoch": 3.97,
      "learning_rate": 2.520801815431165e-05,
      "loss": 0.1663,
      "step": 5247
    },
    {
      "epoch": 3.97,
      "learning_rate": 2.5203290468986384e-05,
      "loss": 0.1549,
      "step": 5248
    },
    {
      "epoch": 3.97,
      "learning_rate": 2.5198562783661123e-05,
      "loss": 0.1427,
      "step": 5249
    },
    {
      "epoch": 3.97,
      "learning_rate": 2.5193835098335855e-05,
      "loss": 0.1488,
      "step": 5250
    },
    {
      "epoch": 3.97,
      "learning_rate": 2.5189107413010593e-05,
      "loss": 0.1489,
      "step": 5251
    },
    {
      "epoch": 3.97,
      "learning_rate": 2.5184379727685326e-05,
      "loss": 0.1335,
      "step": 5252
    },
    {
      "epoch": 3.97,
      "learning_rate": 2.5179652042360064e-05,
      "loss": 0.1486,
      "step": 5253
    },
    {
      "epoch": 3.97,
      "learning_rate": 2.5174924357034796e-05,
      "loss": 0.1431,
      "step": 5254
    },
    {
      "epoch": 3.97,
      "learning_rate": 2.5170196671709535e-05,
      "loss": 0.1535,
      "step": 5255
    },
    {
      "epoch": 3.98,
      "learning_rate": 2.5165468986384267e-05,
      "loss": 0.1519,
      "step": 5256
    },
    {
      "epoch": 3.98,
      "learning_rate": 2.5160741301059003e-05,
      "loss": 0.1436,
      "step": 5257
    },
    {
      "epoch": 3.98,
      "learning_rate": 2.5156013615733738e-05,
      "loss": 0.1371,
      "step": 5258
    },
    {
      "epoch": 3.98,
      "learning_rate": 2.5151285930408474e-05,
      "loss": 0.1703,
      "step": 5259
    },
    {
      "epoch": 3.98,
      "learning_rate": 2.5146558245083206e-05,
      "loss": 0.1505,
      "step": 5260
    },
    {
      "epoch": 3.98,
      "learning_rate": 2.5141830559757945e-05,
      "loss": 0.1408,
      "step": 5261
    },
    {
      "epoch": 3.98,
      "learning_rate": 2.5137102874432677e-05,
      "loss": 0.1504,
      "step": 5262
    },
    {
      "epoch": 3.98,
      "learning_rate": 2.5132375189107415e-05,
      "loss": 0.145,
      "step": 5263
    },
    {
      "epoch": 3.98,
      "learning_rate": 2.5127647503782148e-05,
      "loss": 0.1327,
      "step": 5264
    },
    {
      "epoch": 3.98,
      "learning_rate": 2.5122919818456886e-05,
      "loss": 0.1252,
      "step": 5265
    },
    {
      "epoch": 3.98,
      "learning_rate": 2.511819213313162e-05,
      "loss": 0.1433,
      "step": 5266
    },
    {
      "epoch": 3.98,
      "learning_rate": 2.5113464447806357e-05,
      "loss": 0.151,
      "step": 5267
    },
    {
      "epoch": 3.98,
      "learning_rate": 2.510873676248109e-05,
      "loss": 0.1406,
      "step": 5268
    },
    {
      "epoch": 3.98,
      "learning_rate": 2.5104009077155828e-05,
      "loss": 0.1465,
      "step": 5269
    },
    {
      "epoch": 3.99,
      "learning_rate": 2.509928139183056e-05,
      "loss": 0.1394,
      "step": 5270
    },
    {
      "epoch": 3.99,
      "learning_rate": 2.5094553706505296e-05,
      "loss": 0.164,
      "step": 5271
    },
    {
      "epoch": 3.99,
      "learning_rate": 2.508982602118003e-05,
      "loss": 0.1248,
      "step": 5272
    },
    {
      "epoch": 3.99,
      "learning_rate": 2.5085098335854767e-05,
      "loss": 0.1958,
      "step": 5273
    },
    {
      "epoch": 3.99,
      "learning_rate": 2.50803706505295e-05,
      "loss": 0.1354,
      "step": 5274
    },
    {
      "epoch": 3.99,
      "learning_rate": 2.5075642965204237e-05,
      "loss": 0.1502,
      "step": 5275
    },
    {
      "epoch": 3.99,
      "learning_rate": 2.507091527987897e-05,
      "loss": 0.1542,
      "step": 5276
    },
    {
      "epoch": 3.99,
      "learning_rate": 2.506618759455371e-05,
      "loss": 0.1298,
      "step": 5277
    },
    {
      "epoch": 3.99,
      "learning_rate": 2.506145990922844e-05,
      "loss": 0.1394,
      "step": 5278
    },
    {
      "epoch": 3.99,
      "learning_rate": 2.505673222390318e-05,
      "loss": 0.1631,
      "step": 5279
    },
    {
      "epoch": 3.99,
      "learning_rate": 2.505200453857791e-05,
      "loss": 0.145,
      "step": 5280
    },
    {
      "epoch": 3.99,
      "learning_rate": 2.504727685325265e-05,
      "loss": 0.1428,
      "step": 5281
    },
    {
      "epoch": 3.99,
      "learning_rate": 2.5042549167927382e-05,
      "loss": 0.1409,
      "step": 5282
    },
    {
      "epoch": 4.0,
      "learning_rate": 2.503782148260212e-05,
      "loss": 0.1504,
      "step": 5283
    },
    {
      "epoch": 4.0,
      "learning_rate": 2.5033093797276853e-05,
      "loss": 0.1305,
      "step": 5284
    },
    {
      "epoch": 4.0,
      "learning_rate": 2.502836611195159e-05,
      "loss": 0.1509,
      "step": 5285
    },
    {
      "epoch": 4.0,
      "learning_rate": 2.5023638426626324e-05,
      "loss": 0.1599,
      "step": 5286
    },
    {
      "epoch": 4.0,
      "learning_rate": 2.501891074130106e-05,
      "loss": 0.1491,
      "step": 5287
    },
    {
      "epoch": 4.0,
      "learning_rate": 2.501418305597579e-05,
      "loss": 0.1474,
      "step": 5288
    },
    {
      "epoch": 4.0,
      "learning_rate": 2.500945537065053e-05,
      "loss": 0.1404,
      "step": 5289
    }
  ],
  "logging_steps": 1,
  "max_steps": 10576,
  "num_train_epochs": 8,
  "save_steps": 32,
  "total_flos": 1.8430546283824742e+18,
  "trial_name": null,
  "trial_params": null
}