{
  "best_metric": null,
  "best_model_checkpoint": null,
  "epoch": 3.0,
  "eval_steps": 500,
  "global_step": 1689,
  "is_hyper_param_search": false,
  "is_local_process_zero": true,
  "is_world_process_zero": true,
  "log_history": [
    {
      "epoch": 0.0,
      "learning_rate": 0.0,
      "loss": 0.9229,
      "step": 1
    },
    {
      "epoch": 0.0,
      "learning_rate": 3.0000000000000004e-07,
      "loss": 0.7591,
      "step": 2
    },
    {
      "epoch": 0.01,
      "learning_rate": 6.000000000000001e-07,
      "loss": 0.7555,
      "step": 3
    },
    {
      "epoch": 0.01,
      "learning_rate": 9e-07,
      "loss": 0.7155,
      "step": 4
    },
    {
      "epoch": 0.01,
      "learning_rate": 1.2000000000000002e-06,
      "loss": 0.6815,
      "step": 5
    },
    {
      "epoch": 0.01,
      "learning_rate": 1.5e-06,
      "loss": 0.6505,
      "step": 6
    },
    {
      "epoch": 0.01,
      "learning_rate": 1.8e-06,
      "loss": 0.5808,
      "step": 7
    },
    {
      "epoch": 0.01,
      "learning_rate": 2.1e-06,
      "loss": 0.6016,
      "step": 8
    },
    {
      "epoch": 0.02,
      "learning_rate": 2.4000000000000003e-06,
      "loss": 0.5805,
      "step": 9
    },
    {
      "epoch": 0.02,
      "learning_rate": 2.7e-06,
      "loss": 0.5634,
      "step": 10
    },
    {
      "epoch": 0.02,
      "learning_rate": 3e-06,
      "loss": 0.563,
      "step": 11
    },
    {
      "epoch": 0.02,
      "learning_rate": 3.3e-06,
      "loss": 0.5584,
      "step": 12
    },
    {
      "epoch": 0.02,
      "learning_rate": 3.6e-06,
      "loss": 0.5985,
      "step": 13
    },
    {
      "epoch": 0.02,
      "learning_rate": 3.9e-06,
      "loss": 0.5701,
      "step": 14
    },
    {
      "epoch": 0.03,
      "learning_rate": 4.2e-06,
      "loss": 0.581,
      "step": 15
    },
    {
      "epoch": 0.03,
      "learning_rate": 4.5e-06,
      "loss": 0.5583,
      "step": 16
    },
    {
      "epoch": 0.03,
      "learning_rate": 4.800000000000001e-06,
      "loss": 0.5464,
      "step": 17
    },
    {
      "epoch": 0.03,
      "learning_rate": 5.1e-06,
      "loss": 0.5283,
      "step": 18
    },
    {
      "epoch": 0.03,
      "learning_rate": 5.4e-06,
      "loss": 0.5393,
      "step": 19
    },
    {
      "epoch": 0.04,
      "learning_rate": 5.7e-06,
      "loss": 0.5583,
      "step": 20
    },
    {
      "epoch": 0.04,
      "learning_rate": 6e-06,
      "loss": 0.5323,
      "step": 21
    },
    {
      "epoch": 0.04,
      "learning_rate": 5.9964050329538645e-06,
      "loss": 0.5669,
      "step": 22
    },
    {
      "epoch": 0.04,
      "learning_rate": 5.99281006590773e-06,
      "loss": 0.5353,
      "step": 23
    },
    {
      "epoch": 0.04,
      "learning_rate": 5.989215098861594e-06,
      "loss": 0.5198,
      "step": 24
    },
    {
      "epoch": 0.04,
      "learning_rate": 5.985620131815458e-06,
      "loss": 0.5225,
      "step": 25
    },
    {
      "epoch": 0.05,
      "learning_rate": 5.9820251647693235e-06,
      "loss": 0.5484,
      "step": 26
    },
    {
      "epoch": 0.05,
      "learning_rate": 5.978430197723188e-06,
      "loss": 0.5469,
      "step": 27
    },
    {
      "epoch": 0.05,
      "learning_rate": 5.974835230677052e-06,
      "loss": 0.5411,
      "step": 28
    },
    {
      "epoch": 0.05,
      "learning_rate": 5.9712402636309165e-06,
      "loss": 0.5394,
      "step": 29
    },
    {
      "epoch": 0.05,
      "learning_rate": 5.967645296584782e-06,
      "loss": 0.5481,
      "step": 30
    },
    {
      "epoch": 0.06,
      "learning_rate": 5.964050329538646e-06,
      "loss": 0.5559,
      "step": 31
    },
    {
      "epoch": 0.06,
      "learning_rate": 5.960455362492511e-06,
      "loss": 0.5317,
      "step": 32
    },
    {
      "epoch": 0.06,
      "learning_rate": 5.956860395446375e-06,
      "loss": 0.5266,
      "step": 33
    },
    {
      "epoch": 0.06,
      "learning_rate": 5.95326542840024e-06,
      "loss": 0.5074,
      "step": 34
    },
    {
      "epoch": 0.06,
      "learning_rate": 5.949670461354104e-06,
      "loss": 0.542,
      "step": 35
    },
    {
      "epoch": 0.06,
      "learning_rate": 5.946075494307969e-06,
      "loss": 0.5562,
      "step": 36
    },
    {
      "epoch": 0.07,
      "learning_rate": 5.942480527261834e-06,
      "loss": 0.5381,
      "step": 37
    },
    {
      "epoch": 0.07,
      "learning_rate": 5.938885560215698e-06,
      "loss": 0.5434,
      "step": 38
    },
    {
      "epoch": 0.07,
      "learning_rate": 5.935290593169563e-06,
      "loss": 0.5447,
      "step": 39
    },
    {
      "epoch": 0.07,
      "learning_rate": 5.9316956261234274e-06,
      "loss": 0.5155,
      "step": 40
    },
    {
      "epoch": 0.07,
      "learning_rate": 5.928100659077292e-06,
      "loss": 0.5184,
      "step": 41
    },
    {
      "epoch": 0.07,
      "learning_rate": 5.924505692031156e-06,
      "loss": 0.5373,
      "step": 42
    },
    {
      "epoch": 0.08,
      "learning_rate": 5.920910724985021e-06,
      "loss": 0.549,
      "step": 43
    },
    {
      "epoch": 0.08,
      "learning_rate": 5.917315757938886e-06,
      "loss": 0.5086,
      "step": 44
    },
    {
      "epoch": 0.08,
      "learning_rate": 5.913720790892751e-06,
      "loss": 0.5366,
      "step": 45
    },
    {
      "epoch": 0.08,
      "learning_rate": 5.910125823846615e-06,
      "loss": 0.5262,
      "step": 46
    },
    {
      "epoch": 0.08,
      "learning_rate": 5.906530856800479e-06,
      "loss": 0.5318,
      "step": 47
    },
    {
      "epoch": 0.09,
      "learning_rate": 5.902935889754344e-06,
      "loss": 0.5292,
      "step": 48
    },
    {
      "epoch": 0.09,
      "learning_rate": 5.899340922708209e-06,
      "loss": 0.5335,
      "step": 49
    },
    {
      "epoch": 0.09,
      "learning_rate": 5.895745955662073e-06,
      "loss": 0.5229,
      "step": 50
    },
    {
      "epoch": 0.09,
      "learning_rate": 5.892150988615938e-06,
      "loss": 0.526,
      "step": 51
    },
    {
      "epoch": 0.09,
      "learning_rate": 5.888556021569802e-06,
      "loss": 0.5172,
      "step": 52
    },
    {
      "epoch": 0.09,
      "learning_rate": 5.884961054523667e-06,
      "loss": 0.5351,
      "step": 53
    },
    {
      "epoch": 0.1,
      "learning_rate": 5.881366087477531e-06,
      "loss": 0.5248,
      "step": 54
    },
    {
      "epoch": 0.1,
      "learning_rate": 5.8777711204313966e-06,
      "loss": 0.5031,
      "step": 55
    },
    {
      "epoch": 0.1,
      "learning_rate": 5.874176153385261e-06,
      "loss": 0.5394,
      "step": 56
    },
    {
      "epoch": 0.1,
      "learning_rate": 5.870581186339125e-06,
      "loss": 0.5122,
      "step": 57
    },
    {
      "epoch": 0.1,
      "learning_rate": 5.8669862192929895e-06,
      "loss": 0.5237,
      "step": 58
    },
    {
      "epoch": 0.1,
      "learning_rate": 5.863391252246855e-06,
      "loss": 0.5229,
      "step": 59
    },
    {
      "epoch": 0.11,
      "learning_rate": 5.859796285200719e-06,
      "loss": 0.5297,
      "step": 60
    },
    {
      "epoch": 0.11,
      "learning_rate": 5.856201318154583e-06,
      "loss": 0.5063,
      "step": 61
    },
    {
      "epoch": 0.11,
      "learning_rate": 5.8526063511084485e-06,
      "loss": 0.4923,
      "step": 62
    },
    {
      "epoch": 0.11,
      "learning_rate": 5.849011384062313e-06,
      "loss": 0.5045,
      "step": 63
    },
    {
      "epoch": 0.11,
      "learning_rate": 5.845416417016178e-06,
      "loss": 0.5167,
      "step": 64
    },
    {
      "epoch": 0.12,
      "learning_rate": 5.841821449970042e-06,
      "loss": 0.5266,
      "step": 65
    },
    {
      "epoch": 0.12,
      "learning_rate": 5.838226482923907e-06,
      "loss": 0.5086,
      "step": 66
    },
    {
      "epoch": 0.12,
      "learning_rate": 5.834631515877771e-06,
      "loss": 0.5646,
      "step": 67
    },
    {
      "epoch": 0.12,
      "learning_rate": 5.831036548831636e-06,
      "loss": 0.5131,
      "step": 68
    },
    {
      "epoch": 0.12,
      "learning_rate": 5.8274415817855005e-06,
      "loss": 0.5443,
      "step": 69
    },
    {
      "epoch": 0.12,
      "learning_rate": 5.823846614739365e-06,
      "loss": 0.5233,
      "step": 70
    },
    {
      "epoch": 0.13,
      "learning_rate": 5.820251647693229e-06,
      "loss": 0.5155,
      "step": 71
    },
    {
      "epoch": 0.13,
      "learning_rate": 5.816656680647094e-06,
      "loss": 0.504,
      "step": 72
    },
    {
      "epoch": 0.13,
      "learning_rate": 5.813061713600959e-06,
      "loss": 0.5143,
      "step": 73
    },
    {
      "epoch": 0.13,
      "learning_rate": 5.809466746554824e-06,
      "loss": 0.4983,
      "step": 74
    },
    {
      "epoch": 0.13,
      "learning_rate": 5.805871779508688e-06,
      "loss": 0.4986,
      "step": 75
    },
    {
      "epoch": 0.13,
      "learning_rate": 5.8022768124625525e-06,
      "loss": 0.5211,
      "step": 76
    },
    {
      "epoch": 0.14,
      "learning_rate": 5.798681845416417e-06,
      "loss": 0.5161,
      "step": 77
    },
    {
      "epoch": 0.14,
      "learning_rate": 5.795086878370282e-06,
      "loss": 0.5029,
      "step": 78
    },
    {
      "epoch": 0.14,
      "learning_rate": 5.791491911324146e-06,
      "loss": 0.533,
      "step": 79
    },
    {
      "epoch": 0.14,
      "learning_rate": 5.787896944278011e-06,
      "loss": 0.5426,
      "step": 80
    },
    {
      "epoch": 0.14,
      "learning_rate": 5.784301977231876e-06,
      "loss": 0.5069,
      "step": 81
    },
    {
      "epoch": 0.15,
      "learning_rate": 5.78070701018574e-06,
      "loss": 0.5485,
      "step": 82
    },
    {
      "epoch": 0.15,
      "learning_rate": 5.777112043139605e-06,
      "loss": 0.5172,
      "step": 83
    },
    {
      "epoch": 0.15,
      "learning_rate": 5.773517076093469e-06,
      "loss": 0.5455,
      "step": 84
    },
    {
      "epoch": 0.15,
      "learning_rate": 5.769922109047334e-06,
      "loss": 0.4776,
      "step": 85
    },
    {
      "epoch": 0.15,
      "learning_rate": 5.766327142001198e-06,
      "loss": 0.5452,
      "step": 86
    },
    {
      "epoch": 0.15,
      "learning_rate": 5.7627321749550635e-06,
      "loss": 0.5015,
      "step": 87
    },
    {
      "epoch": 0.16,
      "learning_rate": 5.759137207908928e-06,
      "loss": 0.5196,
      "step": 88
    },
    {
      "epoch": 0.16,
      "learning_rate": 5.755542240862792e-06,
      "loss": 0.5074,
      "step": 89
    },
    {
      "epoch": 0.16,
      "learning_rate": 5.7519472738166564e-06,
      "loss": 0.4911,
      "step": 90
    },
    {
      "epoch": 0.16,
      "learning_rate": 5.748352306770522e-06,
      "loss": 0.4974,
      "step": 91
    },
    {
      "epoch": 0.16,
      "learning_rate": 5.744757339724386e-06,
      "loss": 0.5116,
      "step": 92
    },
    {
      "epoch": 0.17,
      "learning_rate": 5.74116237267825e-06,
      "loss": 0.4898,
      "step": 93
    },
    {
      "epoch": 0.17,
      "learning_rate": 5.7375674056321154e-06,
      "loss": 0.5065,
      "step": 94
    },
    {
      "epoch": 0.17,
      "learning_rate": 5.73397243858598e-06,
      "loss": 0.54,
      "step": 95
    },
    {
      "epoch": 0.17,
      "learning_rate": 5.730377471539844e-06,
      "loss": 0.5341,
      "step": 96
    },
    {
      "epoch": 0.17,
      "learning_rate": 5.726782504493709e-06,
      "loss": 0.5095,
      "step": 97
    },
    {
      "epoch": 0.17,
      "learning_rate": 5.723187537447574e-06,
      "loss": 0.4894,
      "step": 98
    },
    {
      "epoch": 0.18,
      "learning_rate": 5.719592570401438e-06,
      "loss": 0.491,
      "step": 99
    },
    {
      "epoch": 0.18,
      "learning_rate": 5.715997603355303e-06,
      "loss": 0.4862,
      "step": 100
    },
    {
      "epoch": 0.18,
      "learning_rate": 5.712402636309167e-06,
      "loss": 0.4942,
      "step": 101
    },
    {
      "epoch": 0.18,
      "learning_rate": 5.708807669263033e-06,
      "loss": 0.5128,
      "step": 102
    },
    {
      "epoch": 0.18,
      "learning_rate": 5.705212702216896e-06,
      "loss": 0.5089,
      "step": 103
    },
    {
      "epoch": 0.18,
      "learning_rate": 5.701617735170761e-06,
      "loss": 0.5187,
      "step": 104
    },
    {
      "epoch": 0.19,
      "learning_rate": 5.6980227681246256e-06,
      "loss": 0.5084,
      "step": 105
    },
    {
      "epoch": 0.19,
      "learning_rate": 5.694427801078491e-06,
      "loss": 0.5039,
      "step": 106
    },
    {
      "epoch": 0.19,
      "learning_rate": 5.690832834032354e-06,
      "loss": 0.5023,
      "step": 107
    },
    {
      "epoch": 0.19,
      "learning_rate": 5.687237866986219e-06,
      "loss": 0.5209,
      "step": 108
    },
    {
      "epoch": 0.19,
      "learning_rate": 5.683642899940084e-06,
      "loss": 0.5398,
      "step": 109
    },
    {
      "epoch": 0.2,
      "learning_rate": 5.680047932893949e-06,
      "loss": 0.4603,
      "step": 110
    },
    {
      "epoch": 0.2,
      "learning_rate": 5.676452965847813e-06,
      "loss": 0.5129,
      "step": 111
    },
    {
      "epoch": 0.2,
      "learning_rate": 5.6728579988016775e-06,
      "loss": 0.5323,
      "step": 112
    },
    {
      "epoch": 0.2,
      "learning_rate": 5.669263031755542e-06,
      "loss": 0.5372,
      "step": 113
    },
    {
      "epoch": 0.2,
      "learning_rate": 5.665668064709407e-06,
      "loss": 0.5323,
      "step": 114
    },
    {
      "epoch": 0.2,
      "learning_rate": 5.662073097663271e-06,
      "loss": 0.504,
      "step": 115
    },
    {
      "epoch": 0.21,
      "learning_rate": 5.6584781306171365e-06,
      "loss": 0.5196,
      "step": 116
    },
    {
      "epoch": 0.21,
      "learning_rate": 5.654883163571001e-06,
      "loss": 0.4924,
      "step": 117
    },
    {
      "epoch": 0.21,
      "learning_rate": 5.651288196524865e-06,
      "loss": 0.5009,
      "step": 118
    },
    {
      "epoch": 0.21,
      "learning_rate": 5.64769322947873e-06,
      "loss": 0.5275,
      "step": 119
    },
    {
      "epoch": 0.21,
      "learning_rate": 5.644098262432595e-06,
      "loss": 0.5115,
      "step": 120
    },
    {
      "epoch": 0.21,
      "learning_rate": 5.640503295386459e-06,
      "loss": 0.5214,
      "step": 121
    },
    {
      "epoch": 0.22,
      "learning_rate": 5.636908328340323e-06,
      "loss": 0.5126,
      "step": 122
    },
    {
      "epoch": 0.22,
      "learning_rate": 5.6333133612941885e-06,
      "loss": 0.4869,
      "step": 123
    },
    {
      "epoch": 0.22,
      "learning_rate": 5.629718394248053e-06,
      "loss": 0.5011,
      "step": 124
    },
    {
      "epoch": 0.22,
      "learning_rate": 5.626123427201918e-06,
      "loss": 0.4977,
      "step": 125
    },
    {
      "epoch": 0.22,
      "learning_rate": 5.6225284601557815e-06,
      "loss": 0.4864,
      "step": 126
    },
    {
      "epoch": 0.23,
      "learning_rate": 5.618933493109647e-06,
      "loss": 0.5267,
      "step": 127
    },
    {
      "epoch": 0.23,
      "learning_rate": 5.615338526063511e-06,
      "loss": 0.509,
      "step": 128
    },
    {
      "epoch": 0.23,
      "learning_rate": 5.611743559017376e-06,
      "loss": 0.5008,
      "step": 129
    },
    {
      "epoch": 0.23,
      "learning_rate": 5.6081485919712405e-06,
      "loss": 0.486,
      "step": 130
    },
    {
      "epoch": 0.23,
      "learning_rate": 5.604553624925105e-06,
      "loss": 0.5118,
      "step": 131
    },
    {
      "epoch": 0.23,
      "learning_rate": 5.600958657878969e-06,
      "loss": 0.4975,
      "step": 132
    },
    {
      "epoch": 0.24,
      "learning_rate": 5.597363690832834e-06,
      "loss": 0.5084,
      "step": 133
    },
    {
      "epoch": 0.24,
      "learning_rate": 5.593768723786699e-06,
      "loss": 0.5228,
      "step": 134
    },
    {
      "epoch": 0.24,
      "learning_rate": 5.590173756740563e-06,
      "loss": 0.4748,
      "step": 135
    },
    {
      "epoch": 0.24,
      "learning_rate": 5.586578789694428e-06,
      "loss": 0.5363,
      "step": 136
    },
    {
      "epoch": 0.24,
      "learning_rate": 5.5829838226482925e-06,
      "loss": 0.5093,
      "step": 137
    },
    {
      "epoch": 0.25,
      "learning_rate": 5.579388855602158e-06,
      "loss": 0.5086,
      "step": 138
    },
    {
      "epoch": 0.25,
      "learning_rate": 5.575793888556022e-06,
      "loss": 0.5245,
      "step": 139
    },
    {
      "epoch": 0.25,
      "learning_rate": 5.572198921509886e-06,
      "loss": 0.5289,
      "step": 140
    },
    {
      "epoch": 0.25,
      "learning_rate": 5.568603954463751e-06,
      "loss": 0.5149,
      "step": 141
    },
    {
      "epoch": 0.25,
      "learning_rate": 5.565008987417616e-06,
      "loss": 0.5001,
      "step": 142
    },
    {
      "epoch": 0.25,
      "learning_rate": 5.56141402037148e-06,
      "loss": 0.4816,
      "step": 143
    },
    {
      "epoch": 0.26,
      "learning_rate": 5.5578190533253444e-06,
      "loss": 0.5222,
      "step": 144
    },
    {
      "epoch": 0.26,
      "learning_rate": 5.554224086279209e-06,
      "loss": 0.5178,
      "step": 145
    },
    {
      "epoch": 0.26,
      "learning_rate": 5.550629119233074e-06,
      "loss": 0.53,
      "step": 146
    },
    {
      "epoch": 0.26,
      "learning_rate": 5.547034152186938e-06,
      "loss": 0.5055,
      "step": 147
    },
    {
      "epoch": 0.26,
      "learning_rate": 5.5434391851408034e-06,
      "loss": 0.4976,
      "step": 148
    },
    {
      "epoch": 0.26,
      "learning_rate": 5.539844218094667e-06,
      "loss": 0.5311,
      "step": 149
    },
    {
      "epoch": 0.27,
      "learning_rate": 5.536249251048532e-06,
      "loss": 0.5513,
      "step": 150
    },
    {
      "epoch": 0.27,
      "learning_rate": 5.5326542840023964e-06,
      "loss": 0.5028,
      "step": 151
    },
    {
      "epoch": 0.27,
      "learning_rate": 5.529059316956262e-06,
      "loss": 0.4915,
      "step": 152
    },
    {
      "epoch": 0.27,
      "learning_rate": 5.525464349910126e-06,
      "loss": 0.5201,
      "step": 153
    },
    {
      "epoch": 0.27,
      "learning_rate": 5.52186938286399e-06,
      "loss": 0.4976,
      "step": 154
    },
    {
      "epoch": 0.28,
      "learning_rate": 5.518274415817855e-06,
      "loss": 0.4995,
      "step": 155
    },
    {
      "epoch": 0.28,
      "learning_rate": 5.51467944877172e-06,
      "loss": 0.4812,
      "step": 156
    },
    {
      "epoch": 0.28,
      "learning_rate": 5.511084481725585e-06,
      "loss": 0.5333,
      "step": 157
    },
    {
      "epoch": 0.28,
      "learning_rate": 5.507489514679448e-06,
      "loss": 0.5085,
      "step": 158
    },
    {
      "epoch": 0.28,
      "learning_rate": 5.5038945476333136e-06,
      "loss": 0.4969,
      "step": 159
    },
    {
      "epoch": 0.28,
      "learning_rate": 5.500299580587178e-06,
      "loss": 0.5115,
      "step": 160
    },
    {
      "epoch": 0.29,
      "learning_rate": 5.496704613541043e-06,
      "loss": 0.5235,
      "step": 161
    },
    {
      "epoch": 0.29,
      "learning_rate": 5.493109646494907e-06,
      "loss": 0.501,
      "step": 162
    },
    {
      "epoch": 0.29,
      "learning_rate": 5.489514679448772e-06,
      "loss": 0.4767,
      "step": 163
    },
    {
      "epoch": 0.29,
      "learning_rate": 5.485919712402636e-06,
      "loss": 0.5356,
      "step": 164
    },
    {
      "epoch": 0.29,
      "learning_rate": 5.482324745356501e-06,
      "loss": 0.5169,
      "step": 165
    },
    {
      "epoch": 0.29,
      "learning_rate": 5.4787297783103656e-06,
      "loss": 0.5095,
      "step": 166
    },
    {
      "epoch": 0.3,
      "learning_rate": 5.475134811264231e-06,
      "loss": 0.5102,
      "step": 167
    },
    {
      "epoch": 0.3,
      "learning_rate": 5.471539844218094e-06,
      "loss": 0.5152,
      "step": 168
    },
    {
      "epoch": 0.3,
      "learning_rate": 5.467944877171959e-06,
      "loss": 0.5144,
      "step": 169
    },
    {
      "epoch": 0.3,
      "learning_rate": 5.464349910125824e-06,
      "loss": 0.5094,
      "step": 170
    },
    {
      "epoch": 0.3,
      "learning_rate": 5.460754943079689e-06,
      "loss": 0.4966,
      "step": 171
    },
    {
      "epoch": 0.31,
      "learning_rate": 5.457159976033553e-06,
      "loss": 0.5369,
      "step": 172
    },
    {
      "epoch": 0.31,
      "learning_rate": 5.4535650089874175e-06,
      "loss": 0.5108,
      "step": 173
    },
    {
      "epoch": 0.31,
      "learning_rate": 5.449970041941283e-06,
      "loss": 0.5016,
      "step": 174
    },
    {
      "epoch": 0.31,
      "learning_rate": 5.446375074895147e-06,
      "loss": 0.4809,
      "step": 175
    },
    {
      "epoch": 0.31,
      "learning_rate": 5.442780107849012e-06,
      "loss": 0.5172,
      "step": 176
    },
    {
      "epoch": 0.31,
      "learning_rate": 5.439185140802876e-06,
      "loss": 0.5293,
      "step": 177
    },
    {
      "epoch": 0.32,
      "learning_rate": 5.435590173756741e-06,
      "loss": 0.504,
      "step": 178
    },
    {
      "epoch": 0.32,
      "learning_rate": 5.431995206710605e-06,
      "loss": 0.5172,
      "step": 179
    },
    {
      "epoch": 0.32,
      "learning_rate": 5.42840023966447e-06,
      "loss": 0.5095,
      "step": 180
    },
    {
      "epoch": 0.32,
      "learning_rate": 5.424805272618334e-06,
      "loss": 0.5142,
      "step": 181
    },
    {
      "epoch": 0.32,
      "learning_rate": 5.421210305572199e-06,
      "loss": 0.5388,
      "step": 182
    },
    {
      "epoch": 0.33,
      "learning_rate": 5.417615338526063e-06,
      "loss": 0.4757,
      "step": 183
    },
    {
      "epoch": 0.33,
      "learning_rate": 5.4140203714799285e-06,
      "loss": 0.4978,
      "step": 184
    },
    {
      "epoch": 0.33,
      "learning_rate": 5.410425404433793e-06,
      "loss": 0.4793,
      "step": 185
    },
    {
      "epoch": 0.33,
      "learning_rate": 5.406830437387657e-06,
      "loss": 0.5079,
      "step": 186
    },
    {
      "epoch": 0.33,
      "learning_rate": 5.4032354703415215e-06,
      "loss": 0.4696,
      "step": 187
    },
    {
      "epoch": 0.33,
      "learning_rate": 5.399640503295387e-06,
      "loss": 0.5044,
      "step": 188
    },
    {
      "epoch": 0.34,
      "learning_rate": 5.396045536249251e-06,
      "loss": 0.5053,
      "step": 189
    },
    {
      "epoch": 0.34,
      "learning_rate": 5.392450569203116e-06,
      "loss": 0.5167,
      "step": 190
    },
    {
      "epoch": 0.34,
      "learning_rate": 5.3888556021569805e-06,
      "loss": 0.5012,
      "step": 191
    },
    {
      "epoch": 0.34,
      "learning_rate": 5.385260635110845e-06,
      "loss": 0.5129,
      "step": 192
    },
    {
      "epoch": 0.34,
      "learning_rate": 5.38166566806471e-06,
      "loss": 0.4819,
      "step": 193
    },
    {
      "epoch": 0.34,
      "learning_rate": 5.378070701018574e-06,
      "loss": 0.4994,
      "step": 194
    },
    {
      "epoch": 0.35,
      "learning_rate": 5.374475733972439e-06,
      "loss": 0.5041,
      "step": 195
    },
    {
      "epoch": 0.35,
      "learning_rate": 5.370880766926303e-06,
      "loss": 0.4841,
      "step": 196
    },
    {
      "epoch": 0.35,
      "learning_rate": 5.367285799880168e-06,
      "loss": 0.5153,
      "step": 197
    },
    {
      "epoch": 0.35,
      "learning_rate": 5.3636908328340325e-06,
      "loss": 0.4955,
      "step": 198
    },
    {
      "epoch": 0.35,
      "learning_rate": 5.360095865787898e-06,
      "loss": 0.4738,
      "step": 199
    },
    {
      "epoch": 0.36,
      "learning_rate": 5.356500898741761e-06,
      "loss": 0.5227,
      "step": 200
    },
    {
      "epoch": 0.36,
      "learning_rate": 5.352905931695626e-06,
      "loss": 0.4975,
      "step": 201
    },
    {
      "epoch": 0.36,
      "learning_rate": 5.349310964649491e-06,
      "loss": 0.4952,
      "step": 202
    },
    {
      "epoch": 0.36,
      "learning_rate": 5.345715997603356e-06,
      "loss": 0.515,
      "step": 203
    },
    {
      "epoch": 0.36,
      "learning_rate": 5.34212103055722e-06,
      "loss": 0.4993,
      "step": 204
    },
    {
      "epoch": 0.36,
      "learning_rate": 5.3385260635110844e-06,
      "loss": 0.5063,
      "step": 205
    },
    {
      "epoch": 0.37,
      "learning_rate": 5.334931096464949e-06,
      "loss": 0.4895,
      "step": 206
    },
    {
      "epoch": 0.37,
      "learning_rate": 5.331336129418814e-06,
      "loss": 0.5019,
      "step": 207
    },
    {
      "epoch": 0.37,
      "learning_rate": 5.327741162372678e-06,
      "loss": 0.4873,
      "step": 208
    },
    {
      "epoch": 0.37,
      "learning_rate": 5.324146195326543e-06,
      "loss": 0.5416,
      "step": 209
    },
    {
      "epoch": 0.37,
      "learning_rate": 5.320551228280408e-06,
      "loss": 0.5023,
      "step": 210
    },
    {
      "epoch": 0.37,
      "learning_rate": 5.316956261234272e-06,
      "loss": 0.504,
      "step": 211
    },
    {
      "epoch": 0.38,
      "learning_rate": 5.313361294188137e-06,
      "loss": 0.5139,
      "step": 212
    },
    {
      "epoch": 0.38,
      "learning_rate": 5.309766327142002e-06,
      "loss": 0.5052,
      "step": 213
    },
    {
      "epoch": 0.38,
      "learning_rate": 5.306171360095866e-06,
      "loss": 0.5271,
      "step": 214
    },
    {
      "epoch": 0.38,
      "learning_rate": 5.30257639304973e-06,
      "loss": 0.4788,
      "step": 215
    },
    {
      "epoch": 0.38,
      "learning_rate": 5.298981426003595e-06,
      "loss": 0.4837,
      "step": 216
    },
    {
      "epoch": 0.39,
      "learning_rate": 5.29538645895746e-06,
      "loss": 0.5215,
      "step": 217
    },
    {
      "epoch": 0.39,
      "learning_rate": 5.291791491911325e-06,
      "loss": 0.5041,
      "step": 218
    },
    {
      "epoch": 0.39,
      "learning_rate": 5.288196524865188e-06,
      "loss": 0.4784,
      "step": 219
    },
    {
      "epoch": 0.39,
      "learning_rate": 5.2846015578190536e-06,
      "loss": 0.5228,
      "step": 220
    },
    {
      "epoch": 0.39,
      "learning_rate": 5.281006590772918e-06,
      "loss": 0.4943,
      "step": 221
    },
    {
      "epoch": 0.39,
      "learning_rate": 5.277411623726783e-06,
      "loss": 0.5341,
      "step": 222
    },
    {
      "epoch": 0.4,
      "learning_rate": 5.2738166566806465e-06,
      "loss": 0.4695,
      "step": 223
    },
    {
      "epoch": 0.4,
      "learning_rate": 5.270221689634512e-06,
      "loss": 0.5101,
      "step": 224
    },
    {
      "epoch": 0.4,
      "learning_rate": 5.266626722588376e-06,
      "loss": 0.4611,
      "step": 225
    },
    {
      "epoch": 0.4,
      "learning_rate": 5.263031755542241e-06,
      "loss": 0.537,
      "step": 226
    },
    {
      "epoch": 0.4,
      "learning_rate": 5.2594367884961055e-06,
      "loss": 0.5155,
      "step": 227
    },
    {
      "epoch": 0.4,
      "learning_rate": 5.25584182144997e-06,
      "loss": 0.4902,
      "step": 228
    },
    {
      "epoch": 0.41,
      "learning_rate": 5.252246854403835e-06,
      "loss": 0.4946,
      "step": 229
    },
    {
      "epoch": 0.41,
      "learning_rate": 5.248651887357699e-06,
      "loss": 0.5033,
      "step": 230
    },
    {
      "epoch": 0.41,
      "learning_rate": 5.2450569203115645e-06,
      "loss": 0.4956,
      "step": 231
    },
    {
      "epoch": 0.41,
      "learning_rate": 5.241461953265428e-06,
      "loss": 0.5244,
      "step": 232
    },
    {
      "epoch": 0.41,
      "learning_rate": 5.237866986219293e-06,
      "loss": 0.4782,
      "step": 233
    },
    {
      "epoch": 0.42,
      "learning_rate": 5.2342720191731575e-06,
      "loss": 0.4817,
      "step": 234
    },
    {
      "epoch": 0.42,
      "learning_rate": 5.230677052127023e-06,
      "loss": 0.5028,
      "step": 235
    },
    {
      "epoch": 0.42,
      "learning_rate": 5.227082085080887e-06,
      "loss": 0.5069,
      "step": 236
    },
    {
      "epoch": 0.42,
      "learning_rate": 5.223487118034751e-06,
      "loss": 0.5317,
      "step": 237
    },
    {
      "epoch": 0.42,
      "learning_rate": 5.219892150988616e-06,
      "loss": 0.5032,
      "step": 238
    },
    {
      "epoch": 0.42,
      "learning_rate": 5.216297183942481e-06,
      "loss": 0.493,
      "step": 239
    },
    {
      "epoch": 0.43,
      "learning_rate": 5.212702216896345e-06,
      "loss": 0.5169,
      "step": 240
    },
    {
      "epoch": 0.43,
      "learning_rate": 5.20910724985021e-06,
      "loss": 0.5045,
      "step": 241
    },
    {
      "epoch": 0.43,
      "learning_rate": 5.205512282804074e-06,
      "loss": 0.5146,
      "step": 242
    },
    {
      "epoch": 0.43,
      "learning_rate": 5.201917315757939e-06,
      "loss": 0.4969,
      "step": 243
    },
    {
      "epoch": 0.43,
      "learning_rate": 5.198322348711803e-06,
      "loss": 0.4775,
      "step": 244
    },
    {
      "epoch": 0.44,
      "learning_rate": 5.1947273816656685e-06,
      "loss": 0.4821,
      "step": 245
    },
    {
      "epoch": 0.44,
      "learning_rate": 5.191132414619533e-06,
      "loss": 0.5013,
      "step": 246
    },
    {
      "epoch": 0.44,
      "learning_rate": 5.187537447573397e-06,
      "loss": 0.4941,
      "step": 247
    },
    {
      "epoch": 0.44,
      "learning_rate": 5.183942480527262e-06,
      "loss": 0.508,
      "step": 248
    },
    {
      "epoch": 0.44,
      "learning_rate": 5.180347513481127e-06,
      "loss": 0.5256,
      "step": 249
    },
    {
      "epoch": 0.44,
      "learning_rate": 5.176752546434992e-06,
      "loss": 0.4734,
      "step": 250
    },
    {
      "epoch": 0.45,
      "learning_rate": 5.173157579388855e-06,
      "loss": 0.5134,
      "step": 251
    },
    {
      "epoch": 0.45,
      "learning_rate": 5.1695626123427205e-06,
      "loss": 0.4926,
      "step": 252
    },
    {
      "epoch": 0.45,
      "learning_rate": 5.165967645296585e-06,
      "loss": 0.5037,
      "step": 253
    },
    {
      "epoch": 0.45,
      "learning_rate": 5.16237267825045e-06,
      "loss": 0.5102,
      "step": 254
    },
    {
      "epoch": 0.45,
      "learning_rate": 5.158777711204314e-06,
      "loss": 0.4969,
      "step": 255
    },
    {
      "epoch": 0.45,
      "learning_rate": 5.155182744158179e-06,
      "loss": 0.507,
      "step": 256
    },
    {
      "epoch": 0.46,
      "learning_rate": 5.151587777112043e-06,
      "loss": 0.5127,
      "step": 257
    },
    {
      "epoch": 0.46,
      "learning_rate": 5.147992810065908e-06,
      "loss": 0.4634,
      "step": 258
    },
    {
      "epoch": 0.46,
      "learning_rate": 5.1443978430197724e-06,
      "loss": 0.5062,
      "step": 259
    },
    {
      "epoch": 0.46,
      "learning_rate": 5.140802875973637e-06,
      "loss": 0.4894,
      "step": 260
    },
    {
      "epoch": 0.46,
      "learning_rate": 5.137207908927501e-06,
      "loss": 0.4967,
      "step": 261
    },
    {
      "epoch": 0.47,
      "learning_rate": 5.133612941881366e-06,
      "loss": 0.4955,
      "step": 262
    },
    {
      "epoch": 0.47,
      "learning_rate": 5.130017974835231e-06,
      "loss": 0.4918,
      "step": 263
    },
    {
      "epoch": 0.47,
      "learning_rate": 5.126423007789096e-06,
      "loss": 0.5135,
      "step": 264
    },
    {
      "epoch": 0.47,
      "learning_rate": 5.12282804074296e-06,
      "loss": 0.5452,
      "step": 265
    },
    {
      "epoch": 0.47,
      "learning_rate": 5.119233073696824e-06,
      "loss": 0.5095,
      "step": 266
    },
    {
      "epoch": 0.47,
      "learning_rate": 5.11563810665069e-06,
      "loss": 0.4706,
      "step": 267
    },
    {
      "epoch": 0.48,
      "learning_rate": 5.112043139604554e-06,
      "loss": 0.4796,
      "step": 268
    },
    {
      "epoch": 0.48,
      "learning_rate": 5.108448172558419e-06,
      "loss": 0.5066,
      "step": 269
    },
    {
      "epoch": 0.48,
      "learning_rate": 5.1048532055122826e-06,
      "loss": 0.512,
      "step": 270
    },
    {
      "epoch": 0.48,
      "learning_rate": 5.101258238466148e-06,
      "loss": 0.4992,
      "step": 271
    },
    {
      "epoch": 0.48,
      "learning_rate": 5.097663271420012e-06,
      "loss": 0.4764,
      "step": 272
    },
    {
      "epoch": 0.48,
      "learning_rate": 5.094068304373877e-06,
      "loss": 0.4852,
      "step": 273
    },
    {
      "epoch": 0.49,
      "learning_rate": 5.090473337327741e-06,
      "loss": 0.4709,
      "step": 274
    },
    {
      "epoch": 0.49,
      "learning_rate": 5.086878370281606e-06,
      "loss": 0.4892,
      "step": 275
    },
    {
      "epoch": 0.49,
      "learning_rate": 5.08328340323547e-06,
      "loss": 0.4951,
      "step": 276
    },
    {
      "epoch": 0.49,
      "learning_rate": 5.079688436189335e-06,
      "loss": 0.4717,
      "step": 277
    },
    {
      "epoch": 0.49,
      "learning_rate": 5.0760934691432e-06,
      "loss": 0.4874,
      "step": 278
    },
    {
      "epoch": 0.5,
      "learning_rate": 5.072498502097064e-06,
      "loss": 0.4847,
      "step": 279
    },
    {
      "epoch": 0.5,
      "learning_rate": 5.068903535050928e-06,
      "loss": 0.5009,
      "step": 280
    },
    {
      "epoch": 0.5,
      "learning_rate": 5.0653085680047935e-06,
      "loss": 0.5244,
      "step": 281
    },
    {
      "epoch": 0.5,
      "learning_rate": 5.061713600958658e-06,
      "loss": 0.4905,
      "step": 282
    },
    {
      "epoch": 0.5,
      "learning_rate": 5.058118633912522e-06,
      "loss": 0.4975,
      "step": 283
    },
    {
      "epoch": 0.5,
      "learning_rate": 5.054523666866387e-06,
      "loss": 0.4882,
      "step": 284
    },
    {
      "epoch": 0.51,
      "learning_rate": 5.050928699820252e-06,
      "loss": 0.5256,
      "step": 285
    },
    {
      "epoch": 0.51,
      "learning_rate": 5.047333732774117e-06,
      "loss": 0.4818,
      "step": 286
    },
    {
      "epoch": 0.51,
      "learning_rate": 5.043738765727981e-06,
      "loss": 0.4779,
      "step": 287
    },
    {
      "epoch": 0.51,
      "learning_rate": 5.0401437986818455e-06,
      "loss": 0.5093,
      "step": 288
    },
    {
      "epoch": 0.51,
      "learning_rate": 5.03654883163571e-06,
      "loss": 0.4778,
      "step": 289
    },
    {
      "epoch": 0.52,
      "learning_rate": 5.032953864589575e-06,
      "loss": 0.5105,
      "step": 290
    },
    {
      "epoch": 0.52,
      "learning_rate": 5.029358897543439e-06,
      "loss": 0.4991,
      "step": 291
    },
    {
      "epoch": 0.52,
      "learning_rate": 5.0257639304973045e-06,
      "loss": 0.5075,
      "step": 292
    },
    {
      "epoch": 0.52,
      "learning_rate": 5.022168963451168e-06,
      "loss": 0.491,
      "step": 293
    },
    {
      "epoch": 0.52,
      "learning_rate": 5.018573996405033e-06,
      "loss": 0.4922,
      "step": 294
    },
    {
      "epoch": 0.52,
      "learning_rate": 5.0149790293588975e-06,
      "loss": 0.4915,
      "step": 295
    },
    {
      "epoch": 0.53,
      "learning_rate": 5.011384062312763e-06,
      "loss": 0.4806,
      "step": 296
    },
    {
      "epoch": 0.53,
      "learning_rate": 5.007789095266626e-06,
      "loss": 0.5022,
      "step": 297
    },
    {
      "epoch": 0.53,
      "learning_rate": 5.004194128220491e-06,
      "loss": 0.4752,
      "step": 298
    },
    {
      "epoch": 0.53,
      "learning_rate": 5.000599161174356e-06,
      "loss": 0.5015,
      "step": 299
    },
    {
      "epoch": 0.53,
      "learning_rate": 4.997004194128221e-06,
      "loss": 0.5004,
      "step": 300
    },
    {
      "epoch": 0.53,
      "learning_rate": 4.993409227082085e-06,
      "loss": 0.5451,
      "step": 301
    },
    {
      "epoch": 0.54,
      "learning_rate": 4.9898142600359495e-06,
      "loss": 0.4991,
      "step": 302
    },
    {
      "epoch": 0.54,
      "learning_rate": 4.986219292989815e-06,
      "loss": 0.4725,
      "step": 303
    },
    {
      "epoch": 0.54,
      "learning_rate": 4.982624325943679e-06,
      "loss": 0.475,
      "step": 304
    },
    {
      "epoch": 0.54,
      "learning_rate": 4.979029358897544e-06,
      "loss": 0.5085,
      "step": 305
    },
    {
      "epoch": 0.54,
      "learning_rate": 4.9754343918514085e-06,
      "loss": 0.4764,
      "step": 306
    },
    {
      "epoch": 0.55,
      "learning_rate": 4.971839424805273e-06,
      "loss": 0.4855,
      "step": 307
    },
    {
      "epoch": 0.55,
      "learning_rate": 4.968244457759137e-06,
      "loss": 0.4773,
      "step": 308
    },
    {
      "epoch": 0.55,
      "learning_rate": 4.964649490713002e-06,
      "loss": 0.5025,
      "step": 309
    },
    {
      "epoch": 0.55,
      "learning_rate": 4.961054523666867e-06,
      "loss": 0.5135,
      "step": 310
    },
    {
      "epoch": 0.55,
      "learning_rate": 4.957459556620731e-06,
      "loss": 0.4944,
      "step": 311
    },
    {
      "epoch": 0.55,
      "learning_rate": 4.953864589574595e-06,
      "loss": 0.4785,
      "step": 312
    },
    {
      "epoch": 0.56,
      "learning_rate": 4.9502696225284604e-06,
      "loss": 0.4955,
      "step": 313
    },
    {
      "epoch": 0.56,
      "learning_rate": 4.946674655482325e-06,
      "loss": 0.4975,
      "step": 314
    },
    {
      "epoch": 0.56,
      "learning_rate": 4.94307968843619e-06,
      "loss": 0.5225,
      "step": 315
    },
    {
      "epoch": 0.56,
      "learning_rate": 4.9394847213900534e-06,
      "loss": 0.506,
      "step": 316
    },
    {
      "epoch": 0.56,
      "learning_rate": 4.935889754343919e-06,
      "loss": 0.4723,
      "step": 317
    },
    {
      "epoch": 0.56,
      "learning_rate": 4.932294787297783e-06,
      "loss": 0.4832,
      "step": 318
    },
    {
      "epoch": 0.57,
      "learning_rate": 4.928699820251648e-06,
      "loss": 0.5078,
      "step": 319
    },
    {
      "epoch": 0.57,
      "learning_rate": 4.9251048532055124e-06,
      "loss": 0.5033,
      "step": 320
    },
    {
      "epoch": 0.57,
      "learning_rate": 4.921509886159377e-06,
      "loss": 0.4895,
      "step": 321
    },
    {
      "epoch": 0.57,
      "learning_rate": 4.917914919113242e-06,
      "loss": 0.4989,
      "step": 322
    },
    {
      "epoch": 0.57,
      "learning_rate": 4.914319952067106e-06,
      "loss": 0.5324,
      "step": 323
    },
    {
      "epoch": 0.58,
      "learning_rate": 4.9107249850209714e-06,
      "loss": 0.491,
      "step": 324
    },
    {
      "epoch": 0.58,
      "learning_rate": 4.907130017974835e-06,
      "loss": 0.4842,
      "step": 325
    },
    {
      "epoch": 0.58,
      "learning_rate": 4.9035350509287e-06,
      "loss": 0.4999,
      "step": 326
    },
    {
      "epoch": 0.58,
      "learning_rate": 4.899940083882564e-06,
      "loss": 0.4861,
      "step": 327
    },
    {
      "epoch": 0.58,
      "learning_rate": 4.8963451168364296e-06,
      "loss": 0.4713,
      "step": 328
    },
    {
      "epoch": 0.58,
      "learning_rate": 4.892750149790294e-06,
      "loss": 0.4813,
      "step": 329
    },
    {
      "epoch": 0.59,
      "learning_rate": 4.889155182744158e-06,
      "loss": 0.4852,
      "step": 330
    },
    {
      "epoch": 0.59,
      "learning_rate": 4.8855602156980226e-06,
      "loss": 0.5048,
      "step": 331
    },
    {
      "epoch": 0.59,
      "learning_rate": 4.881965248651888e-06,
      "loss": 0.4889,
      "step": 332
    },
    {
      "epoch": 0.59,
      "learning_rate": 4.878370281605752e-06,
      "loss": 0.5085,
      "step": 333
    },
    {
      "epoch": 0.59,
      "learning_rate": 4.874775314559616e-06,
      "loss": 0.5087,
      "step": 334
    },
    {
      "epoch": 0.6,
      "learning_rate": 4.871180347513481e-06,
      "loss": 0.5106,
      "step": 335
    },
    {
      "epoch": 0.6,
      "learning_rate": 4.867585380467346e-06,
      "loss": 0.4768,
      "step": 336
    },
    {
      "epoch": 0.6,
      "learning_rate": 4.86399041342121e-06,
      "loss": 0.4941,
      "step": 337
    },
    {
      "epoch": 0.6,
      "learning_rate": 4.860395446375075e-06,
      "loss": 0.4809,
      "step": 338
    },
    {
      "epoch": 0.6,
      "learning_rate": 4.85680047932894e-06,
      "loss": 0.4988,
      "step": 339
    },
    {
      "epoch": 0.6,
      "learning_rate": 4.853205512282804e-06,
      "loss": 0.532,
      "step": 340
    },
    {
      "epoch": 0.61,
      "learning_rate": 4.849610545236669e-06,
      "loss": 0.4794,
      "step": 341
    },
    {
      "epoch": 0.61,
      "learning_rate": 4.8460155781905335e-06,
      "loss": 0.4897,
      "step": 342
    },
    {
      "epoch": 0.61,
      "learning_rate": 4.842420611144398e-06,
      "loss": 0.5108,
      "step": 343
    },
    {
      "epoch": 0.61,
      "learning_rate": 4.838825644098262e-06,
      "loss": 0.4896,
      "step": 344
    },
    {
      "epoch": 0.61,
      "learning_rate": 4.835230677052127e-06,
      "loss": 0.4659,
      "step": 345
    },
    {
      "epoch": 0.61,
      "learning_rate": 4.831635710005992e-06,
      "loss": 0.4569,
      "step": 346
    },
    {
      "epoch": 0.62,
      "learning_rate": 4.828040742959857e-06,
      "loss": 0.4807,
      "step": 347
    },
    {
      "epoch": 0.62,
      "learning_rate": 4.82444577591372e-06,
      "loss": 0.4759,
      "step": 348
    },
    {
      "epoch": 0.62,
      "learning_rate": 4.8208508088675855e-06,
      "loss": 0.4879,
      "step": 349
    },
    {
      "epoch": 0.62,
      "learning_rate": 4.81725584182145e-06,
      "loss": 0.4903,
      "step": 350
    },
    {
      "epoch": 0.62,
      "learning_rate": 4.813660874775315e-06,
      "loss": 0.5096,
      "step": 351
    },
    {
      "epoch": 0.63,
      "learning_rate": 4.810065907729179e-06,
      "loss": 0.4695,
      "step": 352
    },
    {
      "epoch": 0.63,
      "learning_rate": 4.806470940683044e-06,
      "loss": 0.4806,
      "step": 353
    },
    {
      "epoch": 0.63,
      "learning_rate": 4.802875973636908e-06,
      "loss": 0.4945,
      "step": 354
    },
    {
      "epoch": 0.63,
      "learning_rate": 4.799281006590773e-06,
      "loss": 0.4978,
      "step": 355
    },
    {
      "epoch": 0.63,
      "learning_rate": 4.7956860395446375e-06,
      "loss": 0.4941,
      "step": 356
    },
    {
      "epoch": 0.63,
      "learning_rate": 4.792091072498503e-06,
      "loss": 0.5129,
      "step": 357
    },
    {
      "epoch": 0.64,
      "learning_rate": 4.788496105452367e-06,
      "loss": 0.4641,
      "step": 358
    },
    {
      "epoch": 0.64,
      "learning_rate": 4.784901138406231e-06,
      "loss": 0.4837,
      "step": 359
    },
    {
      "epoch": 0.64,
      "learning_rate": 4.7813061713600965e-06,
      "loss": 0.4905,
      "step": 360
    },
    {
      "epoch": 0.64,
      "learning_rate": 4.777711204313961e-06,
      "loss": 0.5069,
      "step": 361
    },
    {
      "epoch": 0.64,
      "learning_rate": 4.774116237267825e-06,
      "loss": 0.4844,
      "step": 362
    },
    {
      "epoch": 0.64,
      "learning_rate": 4.7705212702216895e-06,
      "loss": 0.4882,
      "step": 363
    },
    {
      "epoch": 0.65,
      "learning_rate": 4.766926303175555e-06,
      "loss": 0.5054,
      "step": 364
    },
    {
      "epoch": 0.65,
      "learning_rate": 4.763331336129419e-06,
      "loss": 0.501,
      "step": 365
    },
    {
      "epoch": 0.65,
      "learning_rate": 4.759736369083284e-06,
      "loss": 0.4767,
      "step": 366
    },
    {
      "epoch": 0.65,
      "learning_rate": 4.756141402037148e-06,
      "loss": 0.4982,
      "step": 367
    },
    {
      "epoch": 0.65,
      "learning_rate": 4.752546434991013e-06,
      "loss": 0.5001,
      "step": 368
    },
    {
      "epoch": 0.66,
      "learning_rate": 4.748951467944877e-06,
      "loss": 0.4883,
      "step": 369
    },
    {
      "epoch": 0.66,
      "learning_rate": 4.745356500898742e-06,
      "loss": 0.5125,
      "step": 370
    },
    {
      "epoch": 0.66,
      "learning_rate": 4.741761533852607e-06,
      "loss": 0.4736,
      "step": 371
    },
    {
      "epoch": 0.66,
      "learning_rate": 4.738166566806471e-06,
      "loss": 0.5074,
      "step": 372
    },
    {
      "epoch": 0.66,
      "learning_rate": 4.734571599760335e-06,
      "loss": 0.4762,
      "step": 373
    },
    {
      "epoch": 0.66,
      "learning_rate": 4.7309766327142004e-06,
      "loss": 0.4984,
      "step": 374
    },
    {
      "epoch": 0.67,
      "learning_rate": 4.727381665668065e-06,
      "loss": 0.4808,
      "step": 375
    },
    {
      "epoch": 0.67,
      "learning_rate": 4.723786698621929e-06,
      "loss": 0.4801,
      "step": 376
    },
    {
      "epoch": 0.67,
      "learning_rate": 4.720191731575794e-06,
      "loss": 0.4886,
      "step": 377
    },
    {
      "epoch": 0.67,
      "learning_rate": 4.716596764529659e-06,
      "loss": 0.5019,
      "step": 378
    },
    {
      "epoch": 0.67,
      "learning_rate": 4.713001797483523e-06,
      "loss": 0.5053,
      "step": 379
    },
    {
      "epoch": 0.67,
      "learning_rate": 4.709406830437388e-06,
      "loss": 0.4908,
      "step": 380
    },
    {
      "epoch": 0.68,
      "learning_rate": 4.705811863391252e-06,
      "loss": 0.5027,
      "step": 381
    },
    {
      "epoch": 0.68,
      "learning_rate": 4.702216896345117e-06,
      "loss": 0.4523,
      "step": 382
    },
    {
      "epoch": 0.68,
      "learning_rate": 4.698621929298982e-06,
      "loss": 0.4745,
      "step": 383
    },
    {
      "epoch": 0.68,
      "learning_rate": 4.695026962252846e-06,
      "loss": 0.5041,
      "step": 384
    },
    {
      "epoch": 0.68,
      "learning_rate": 4.6914319952067106e-06,
      "loss": 0.495,
      "step": 385
    },
    {
      "epoch": 0.69,
      "learning_rate": 4.687837028160575e-06,
      "loss": 0.5137,
      "step": 386
    },
    {
      "epoch": 0.69,
      "learning_rate": 4.68424206111444e-06,
      "loss": 0.4938,
      "step": 387
    },
    {
      "epoch": 0.69,
      "learning_rate": 4.680647094068304e-06,
      "loss": 0.4915,
      "step": 388
    },
    {
      "epoch": 0.69,
      "learning_rate": 4.6770521270221696e-06,
      "loss": 0.4986,
      "step": 389
    },
    {
      "epoch": 0.69,
      "learning_rate": 4.673457159976033e-06,
      "loss": 0.4995,
      "step": 390
    },
    {
      "epoch": 0.69,
      "learning_rate": 4.669862192929898e-06,
      "loss": 0.4721,
      "step": 391
    },
    {
      "epoch": 0.7,
      "learning_rate": 4.6662672258837625e-06,
      "loss": 0.5002,
      "step": 392
    },
    {
      "epoch": 0.7,
      "learning_rate": 4.662672258837628e-06,
      "loss": 0.4997,
      "step": 393
    },
    {
      "epoch": 0.7,
      "learning_rate": 4.659077291791492e-06,
      "loss": 0.4821,
      "step": 394
    },
    {
      "epoch": 0.7,
      "learning_rate": 4.655482324745356e-06,
      "loss": 0.5035,
      "step": 395
    },
    {
      "epoch": 0.7,
      "learning_rate": 4.6518873576992215e-06,
      "loss": 0.5115,
      "step": 396
    },
    {
      "epoch": 0.71,
      "learning_rate": 4.648292390653086e-06,
      "loss": 0.4669,
      "step": 397
    },
    {
      "epoch": 0.71,
      "learning_rate": 4.64469742360695e-06,
      "loss": 0.4968,
      "step": 398
    },
    {
      "epoch": 0.71,
      "learning_rate": 4.6411024565608145e-06,
      "loss": 0.4757,
      "step": 399
    },
    {
      "epoch": 0.71,
      "learning_rate": 4.63750748951468e-06,
      "loss": 0.5114,
      "step": 400
    },
    {
      "epoch": 0.71,
      "learning_rate": 4.633912522468544e-06,
      "loss": 0.4795,
      "step": 401
    },
    {
      "epoch": 0.71,
      "learning_rate": 4.630317555422409e-06,
      "loss": 0.4783,
      "step": 402
    },
    {
      "epoch": 0.72,
      "learning_rate": 4.6267225883762735e-06,
      "loss": 0.4855,
      "step": 403
    },
    {
      "epoch": 0.72,
      "learning_rate": 4.623127621330138e-06,
      "loss": 0.4748,
      "step": 404
    },
    {
      "epoch": 0.72,
      "learning_rate": 4.619532654284002e-06,
      "loss": 0.5073,
      "step": 405
    },
    {
      "epoch": 0.72,
      "learning_rate": 4.615937687237867e-06,
      "loss": 0.5255,
      "step": 406
    },
    {
      "epoch": 0.72,
      "learning_rate": 4.612342720191732e-06,
      "loss": 0.4903,
      "step": 407
    },
    {
      "epoch": 0.72,
      "learning_rate": 4.608747753145597e-06,
      "loss": 0.4922,
      "step": 408
    },
    {
      "epoch": 0.73,
      "learning_rate": 4.60515278609946e-06,
      "loss": 0.4746,
      "step": 409
    },
    {
      "epoch": 0.73,
      "learning_rate": 4.6015578190533255e-06,
      "loss": 0.5177,
      "step": 410
    },
    {
      "epoch": 0.73,
      "learning_rate": 4.59796285200719e-06,
      "loss": 0.5198,
      "step": 411
    },
    {
      "epoch": 0.73,
      "learning_rate": 4.594367884961055e-06,
      "loss": 0.5077,
      "step": 412
    },
    {
      "epoch": 0.73,
      "learning_rate": 4.590772917914919e-06,
      "loss": 0.473,
      "step": 413
    },
    {
      "epoch": 0.74,
      "learning_rate": 4.587177950868784e-06,
      "loss": 0.5046,
      "step": 414
    },
    {
      "epoch": 0.74,
      "learning_rate": 4.583582983822649e-06,
      "loss": 0.5046,
      "step": 415
    },
    {
      "epoch": 0.74,
      "learning_rate": 4.579988016776513e-06,
      "loss": 0.4942,
      "step": 416
    },
    {
      "epoch": 0.74,
      "learning_rate": 4.5763930497303775e-06,
      "loss": 0.4893,
      "step": 417
    },
    {
      "epoch": 0.74,
      "learning_rate": 4.572798082684242e-06,
      "loss": 0.4883,
      "step": 418
    },
    {
      "epoch": 0.74,
      "learning_rate": 4.569203115638107e-06,
      "loss": 0.5128,
      "step": 419
    },
    {
      "epoch": 0.75,
      "learning_rate": 4.565608148591971e-06,
      "loss": 0.4822,
      "step": 420
    },
    {
      "epoch": 0.75,
      "learning_rate": 4.5620131815458365e-06,
      "loss": 0.4791,
      "step": 421
    },
    {
      "epoch": 0.75,
      "learning_rate": 4.558418214499701e-06,
      "loss": 0.4976,
      "step": 422
    },
    {
      "epoch": 0.75,
      "learning_rate": 4.554823247453565e-06,
      "loss": 0.4866,
      "step": 423
    },
    {
      "epoch": 0.75,
      "learning_rate": 4.5512282804074294e-06,
      "loss": 0.4893,
      "step": 424
    },
    {
      "epoch": 0.75,
      "learning_rate": 4.547633313361295e-06,
      "loss": 0.4833,
      "step": 425
    },
    {
      "epoch": 0.76,
      "learning_rate": 4.544038346315159e-06,
      "loss": 0.4658,
      "step": 426
    },
    {
      "epoch": 0.76,
      "learning_rate": 4.540443379269023e-06,
      "loss": 0.5027,
      "step": 427
    },
    {
      "epoch": 0.76,
      "learning_rate": 4.536848412222888e-06,
      "loss": 0.4968,
      "step": 428
    },
    {
      "epoch": 0.76,
      "learning_rate": 4.533253445176753e-06,
      "loss": 0.487,
      "step": 429
    },
    {
      "epoch": 0.76,
      "learning_rate": 4.529658478130617e-06,
      "loss": 0.5291,
      "step": 430
    },
    {
      "epoch": 0.77,
      "learning_rate": 4.526063511084482e-06,
      "loss": 0.4751,
      "step": 431
    },
    {
      "epoch": 0.77,
      "learning_rate": 4.522468544038347e-06,
      "loss": 0.4994,
      "step": 432
    },
    {
      "epoch": 0.77,
      "learning_rate": 4.518873576992211e-06,
      "loss": 0.4601,
      "step": 433
    },
    {
      "epoch": 0.77,
      "learning_rate": 4.515278609946075e-06,
      "loss": 0.4834,
      "step": 434
    },
    {
      "epoch": 0.77,
      "learning_rate": 4.51168364289994e-06,
      "loss": 0.4551,
      "step": 435
    },
    {
      "epoch": 0.77,
      "learning_rate": 4.508088675853805e-06,
      "loss": 0.5271,
      "step": 436
    },
    {
      "epoch": 0.78,
      "learning_rate": 4.504493708807669e-06,
      "loss": 0.4846,
      "step": 437
    },
    {
      "epoch": 0.78,
      "learning_rate": 4.500898741761534e-06,
      "loss": 0.4867,
      "step": 438
    },
    {
      "epoch": 0.78,
      "learning_rate": 4.4973037747153986e-06,
      "loss": 0.476,
      "step": 439
    },
    {
      "epoch": 0.78,
      "learning_rate": 4.493708807669264e-06,
      "loss": 0.5,
      "step": 440
    },
    {
      "epoch": 0.78,
      "learning_rate": 4.490113840623127e-06,
      "loss": 0.5228,
      "step": 441
    },
    {
      "epoch": 0.79,
      "learning_rate": 4.486518873576992e-06,
      "loss": 0.4965,
      "step": 442
    },
    {
      "epoch": 0.79,
      "learning_rate": 4.482923906530857e-06,
      "loss": 0.4782,
      "step": 443
    },
    {
      "epoch": 0.79,
      "learning_rate": 4.479328939484722e-06,
      "loss": 0.5109,
      "step": 444
    },
    {
      "epoch": 0.79,
      "learning_rate": 4.475733972438586e-06,
      "loss": 0.4992,
      "step": 445
    },
    {
      "epoch": 0.79,
      "learning_rate": 4.4721390053924505e-06,
      "loss": 0.5033,
      "step": 446
    },
    {
      "epoch": 0.79,
      "learning_rate": 4.468544038346315e-06,
      "loss": 0.4931,
      "step": 447
    },
    {
      "epoch": 0.8,
      "learning_rate": 4.46494907130018e-06,
      "loss": 0.4994,
      "step": 448
    },
    {
      "epoch": 0.8,
      "learning_rate": 4.461354104254044e-06,
      "loss": 0.4809,
      "step": 449
    },
    {
      "epoch": 0.8,
      "learning_rate": 4.457759137207909e-06,
      "loss": 0.4824,
      "step": 450
    },
    {
      "epoch": 0.8,
      "learning_rate": 4.454164170161774e-06,
      "loss": 0.5227,
      "step": 451
    },
    {
      "epoch": 0.8,
      "learning_rate": 4.450569203115638e-06,
      "loss": 0.4839,
      "step": 452
    },
    {
      "epoch": 0.8,
      "learning_rate": 4.4469742360695025e-06,
      "loss": 0.5037,
      "step": 453
    },
    {
      "epoch": 0.81,
      "learning_rate": 4.443379269023368e-06,
      "loss": 0.4911,
      "step": 454
    },
    {
      "epoch": 0.81,
      "learning_rate": 4.439784301977232e-06,
      "loss": 0.4992,
      "step": 455
    },
    {
      "epoch": 0.81,
      "learning_rate": 4.436189334931096e-06,
      "loss": 0.506,
      "step": 456
    },
    {
      "epoch": 0.81,
      "learning_rate": 4.4325943678849615e-06,
      "loss": 0.5018,
      "step": 457
    },
    {
      "epoch": 0.81,
      "learning_rate": 4.428999400838826e-06,
      "loss": 0.4862,
      "step": 458
    },
    {
      "epoch": 0.82,
      "learning_rate": 4.425404433792691e-06,
      "loss": 0.4935,
      "step": 459
    },
    {
      "epoch": 0.82,
      "learning_rate": 4.4218094667465545e-06,
      "loss": 0.5005,
      "step": 460
    },
    {
      "epoch": 0.82,
      "learning_rate": 4.41821449970042e-06,
      "loss": 0.479,
      "step": 461
    },
    {
      "epoch": 0.82,
      "learning_rate": 4.414619532654284e-06,
      "loss": 0.4998,
      "step": 462
    },
    {
      "epoch": 0.82,
      "learning_rate": 4.411024565608149e-06,
      "loss": 0.4709,
      "step": 463
    },
    {
      "epoch": 0.82,
      "learning_rate": 4.407429598562013e-06,
      "loss": 0.4692,
      "step": 464
    },
    {
      "epoch": 0.83,
      "learning_rate": 4.403834631515878e-06,
      "loss": 0.4906,
      "step": 465
    },
    {
      "epoch": 0.83,
      "learning_rate": 4.400239664469742e-06,
      "loss": 0.4811,
      "step": 466
    },
    {
      "epoch": 0.83,
      "learning_rate": 4.396644697423607e-06,
      "loss": 0.4752,
      "step": 467
    },
    {
      "epoch": 0.83,
      "learning_rate": 4.393049730377472e-06,
      "loss": 0.4853,
      "step": 468
    },
    {
      "epoch": 0.83,
      "learning_rate": 4.389454763331336e-06,
      "loss": 0.4925,
      "step": 469
    },
    {
      "epoch": 0.83,
      "learning_rate": 4.385859796285201e-06,
      "loss": 0.4842,
      "step": 470
    },
    {
      "epoch": 0.84,
      "learning_rate": 4.3822648292390655e-06,
      "loss": 0.5173,
      "step": 471
    },
    {
      "epoch": 0.84,
      "learning_rate": 4.37866986219293e-06,
      "loss": 0.4944,
      "step": 472
    },
    {
      "epoch": 0.84,
      "learning_rate": 4.375074895146795e-06,
      "loss": 0.4875,
      "step": 473
    },
    {
      "epoch": 0.84,
      "learning_rate": 4.371479928100659e-06,
      "loss": 0.488,
      "step": 474
    },
    {
      "epoch": 0.84,
      "learning_rate": 4.367884961054524e-06,
      "loss": 0.4868,
      "step": 475
    },
    {
      "epoch": 0.85,
      "learning_rate": 4.364289994008389e-06,
      "loss": 0.4846,
      "step": 476
    },
    {
      "epoch": 0.85,
      "learning_rate": 4.360695026962253e-06,
      "loss": 0.4968,
      "step": 477
    },
    {
      "epoch": 0.85,
      "learning_rate": 4.3571000599161175e-06,
      "loss": 0.4941,
      "step": 478
    },
    {
      "epoch": 0.85,
      "learning_rate": 4.353505092869982e-06,
      "loss": 0.4791,
      "step": 479
    },
    {
      "epoch": 0.85,
      "learning_rate": 4.349910125823847e-06,
      "loss": 0.5,
      "step": 480
    },
    {
      "epoch": 0.85,
      "learning_rate": 4.346315158777711e-06,
      "loss": 0.5054,
      "step": 481
    },
    {
      "epoch": 0.86,
      "learning_rate": 4.3427201917315765e-06,
      "loss": 0.4615,
      "step": 482
    },
    {
      "epoch": 0.86,
      "learning_rate": 4.33912522468544e-06,
      "loss": 0.5011,
      "step": 483
    },
    {
      "epoch": 0.86,
      "learning_rate": 4.335530257639305e-06,
      "loss": 0.4855,
      "step": 484
    },
    {
      "epoch": 0.86,
      "learning_rate": 4.3319352905931694e-06,
      "loss": 0.473,
      "step": 485
    },
    {
      "epoch": 0.86,
      "learning_rate": 4.328340323547035e-06,
      "loss": 0.4815,
      "step": 486
    },
    {
      "epoch": 0.87,
      "learning_rate": 4.324745356500899e-06,
      "loss": 0.5057,
      "step": 487
    },
    {
      "epoch": 0.87,
      "learning_rate": 4.321150389454763e-06,
      "loss": 0.4494,
      "step": 488
    },
    {
      "epoch": 0.87,
      "learning_rate": 4.317555422408628e-06,
      "loss": 0.4738,
      "step": 489
    },
    {
      "epoch": 0.87,
      "learning_rate": 4.313960455362493e-06,
      "loss": 0.4973,
      "step": 490
    },
    {
      "epoch": 0.87,
      "learning_rate": 4.310365488316357e-06,
      "loss": 0.504,
      "step": 491
    },
    {
      "epoch": 0.87,
      "learning_rate": 4.306770521270221e-06,
      "loss": 0.4849,
      "step": 492
    },
    {
      "epoch": 0.88,
      "learning_rate": 4.303175554224087e-06,
      "loss": 0.4936,
      "step": 493
    },
    {
      "epoch": 0.88,
      "learning_rate": 4.299580587177951e-06,
      "loss": 0.4821,
      "step": 494
    },
    {
      "epoch": 0.88,
      "learning_rate": 4.295985620131816e-06,
      "loss": 0.467,
      "step": 495
    },
    {
      "epoch": 0.88,
      "learning_rate": 4.29239065308568e-06,
      "loss": 0.4891,
      "step": 496
    },
    {
      "epoch": 0.88,
      "learning_rate": 4.288795686039545e-06,
      "loss": 0.4996,
      "step": 497
    },
    {
      "epoch": 0.88,
      "learning_rate": 4.285200718993409e-06,
      "loss": 0.458,
      "step": 498
    },
    {
      "epoch": 0.89,
      "learning_rate": 4.281605751947274e-06,
      "loss": 0.4897,
      "step": 499
    },
    {
      "epoch": 0.89,
      "learning_rate": 4.2780107849011386e-06,
      "loss": 0.4938,
      "step": 500
    },
    {
      "epoch": 0.89,
      "learning_rate": 4.274415817855003e-06,
      "loss": 0.4874,
      "step": 501
    },
    {
      "epoch": 0.89,
      "learning_rate": 4.270820850808867e-06,
      "loss": 0.5215,
      "step": 502
    },
    {
      "epoch": 0.89,
      "learning_rate": 4.267225883762732e-06,
      "loss": 0.5014,
      "step": 503
    },
    {
      "epoch": 0.9,
      "learning_rate": 4.263630916716597e-06,
      "loss": 0.4803,
      "step": 504
    },
    {
      "epoch": 0.9,
      "learning_rate": 4.260035949670462e-06,
      "loss": 0.5022,
      "step": 505
    },
    {
      "epoch": 0.9,
      "learning_rate": 4.256440982624326e-06,
      "loss": 0.5074,
      "step": 506
    },
    {
      "epoch": 0.9,
      "learning_rate": 4.2528460155781905e-06,
      "loss": 0.4959,
      "step": 507
    },
    {
      "epoch": 0.9,
      "learning_rate": 4.249251048532055e-06,
      "loss": 0.5096,
      "step": 508
    },
    {
      "epoch": 0.9,
      "learning_rate": 4.24565608148592e-06,
      "loss": 0.4762,
      "step": 509
    },
    {
      "epoch": 0.91,
      "learning_rate": 4.242061114439784e-06,
      "loss": 0.5014,
      "step": 510
    },
    {
      "epoch": 0.91,
      "learning_rate": 4.238466147393649e-06,
      "loss": 0.4846,
      "step": 511
    },
    {
      "epoch": 0.91,
      "learning_rate": 4.234871180347514e-06,
      "loss": 0.4862,
      "step": 512
    },
    {
      "epoch": 0.91,
      "learning_rate": 4.231276213301378e-06,
      "loss": 0.4589,
      "step": 513
    },
    {
      "epoch": 0.91,
      "learning_rate": 4.227681246255243e-06,
      "loss": 0.4955,
      "step": 514
    },
    {
      "epoch": 0.91,
      "learning_rate": 4.224086279209107e-06,
      "loss": 0.4754,
      "step": 515
    },
    {
      "epoch": 0.92,
      "learning_rate": 4.220491312162972e-06,
      "loss": 0.4998,
      "step": 516
    },
    {
      "epoch": 0.92,
      "learning_rate": 4.216896345116836e-06,
      "loss": 0.475,
      "step": 517
    },
    {
      "epoch": 0.92,
      "learning_rate": 4.2133013780707015e-06,
      "loss": 0.5176,
      "step": 518
    },
    {
      "epoch": 0.92,
      "learning_rate": 4.209706411024566e-06,
      "loss": 0.5194,
      "step": 519
    },
    {
      "epoch": 0.92,
      "learning_rate": 4.20611144397843e-06,
      "loss": 0.5138,
      "step": 520
    },
    {
      "epoch": 0.93,
      "learning_rate": 4.2025164769322945e-06,
      "loss": 0.4778,
      "step": 521
    },
    {
      "epoch": 0.93,
      "learning_rate": 4.19892150988616e-06,
      "loss": 0.4754,
      "step": 522
    },
    {
      "epoch": 0.93,
      "learning_rate": 4.195326542840024e-06,
      "loss": 0.4905,
      "step": 523
    },
    {
      "epoch": 0.93,
      "learning_rate": 4.191731575793889e-06,
      "loss": 0.4896,
      "step": 524
    },
    {
      "epoch": 0.93,
      "learning_rate": 4.188136608747753e-06,
      "loss": 0.4694,
      "step": 525
    },
    {
      "epoch": 0.93,
      "learning_rate": 4.184541641701618e-06,
      "loss": 0.5051,
      "step": 526
    },
    {
      "epoch": 0.94,
      "learning_rate": 4.180946674655482e-06,
      "loss": 0.4834,
      "step": 527
    },
    {
      "epoch": 0.94,
      "learning_rate": 4.177351707609347e-06,
      "loss": 0.493,
      "step": 528
    },
    {
      "epoch": 0.94,
      "learning_rate": 4.173756740563212e-06,
      "loss": 0.5016,
      "step": 529
    },
    {
      "epoch": 0.94,
      "learning_rate": 4.170161773517076e-06,
      "loss": 0.504,
      "step": 530
    },
    {
      "epoch": 0.94,
      "learning_rate": 4.166566806470941e-06,
      "loss": 0.4803,
      "step": 531
    },
    {
      "epoch": 0.94,
      "learning_rate": 4.1629718394248055e-06,
      "loss": 0.5005,
      "step": 532
    },
    {
      "epoch": 0.95,
      "learning_rate": 4.159376872378671e-06,
      "loss": 0.4765,
      "step": 533
    },
    {
      "epoch": 0.95,
      "learning_rate": 4.155781905332534e-06,
      "loss": 0.4913,
      "step": 534
    },
    {
      "epoch": 0.95,
      "learning_rate": 4.152186938286399e-06,
      "loss": 0.4876,
      "step": 535
    },
    {
      "epoch": 0.95,
      "learning_rate": 4.148591971240264e-06,
      "loss": 0.488,
      "step": 536
    },
    {
      "epoch": 0.95,
      "learning_rate": 4.144997004194129e-06,
      "loss": 0.4774,
      "step": 537
    },
    {
      "epoch": 0.96,
      "learning_rate": 4.141402037147992e-06,
      "loss": 0.4989,
      "step": 538
    },
    {
      "epoch": 0.96,
      "learning_rate": 4.1378070701018574e-06,
      "loss": 0.4901,
      "step": 539
    },
    {
      "epoch": 0.96,
      "learning_rate": 4.134212103055722e-06,
      "loss": 0.4861,
      "step": 540
    },
    {
      "epoch": 0.96,
      "learning_rate": 4.130617136009587e-06,
      "loss": 0.5002,
      "step": 541
    },
    {
      "epoch": 0.96,
      "learning_rate": 4.127022168963451e-06,
      "loss": 0.5004,
      "step": 542
    },
    {
      "epoch": 0.96,
      "learning_rate": 4.123427201917316e-06,
      "loss": 0.4614,
      "step": 543
    },
    {
      "epoch": 0.97,
      "learning_rate": 4.11983223487118e-06,
      "loss": 0.4672,
      "step": 544
    },
    {
      "epoch": 0.97,
      "learning_rate": 4.116237267825045e-06,
      "loss": 0.4754,
      "step": 545
    },
    {
      "epoch": 0.97,
      "learning_rate": 4.112642300778909e-06,
      "loss": 0.4927,
      "step": 546
    },
    {
      "epoch": 0.97,
      "learning_rate": 4.109047333732775e-06,
      "loss": 0.4749,
      "step": 547
    },
    {
      "epoch": 0.97,
      "learning_rate": 4.105452366686639e-06,
      "loss": 0.4858,
      "step": 548
    },
    {
      "epoch": 0.98,
      "learning_rate": 4.101857399640503e-06,
      "loss": 0.4632,
      "step": 549
    },
    {
      "epoch": 0.98,
      "learning_rate": 4.098262432594368e-06,
      "loss": 0.4892,
      "step": 550
    },
    {
      "epoch": 0.98,
      "learning_rate": 4.094667465548233e-06,
      "loss": 0.5137,
      "step": 551
    },
    {
      "epoch": 0.98,
      "learning_rate": 4.091072498502097e-06,
      "loss": 0.4757,
      "step": 552
    },
    {
      "epoch": 0.98,
      "learning_rate": 4.087477531455961e-06,
      "loss": 0.4789,
      "step": 553
    },
    {
      "epoch": 0.98,
      "learning_rate": 4.0838825644098266e-06,
      "loss": 0.468,
      "step": 554
    },
    {
      "epoch": 0.99,
      "learning_rate": 4.080287597363691e-06,
      "loss": 0.4887,
      "step": 555
    },
    {
      "epoch": 0.99,
      "learning_rate": 4.076692630317556e-06,
      "loss": 0.4563,
      "step": 556
    },
    {
      "epoch": 0.99,
      "learning_rate": 4.0730976632714195e-06,
      "loss": 0.5165,
      "step": 557
    },
    {
      "epoch": 0.99,
      "learning_rate": 4.069502696225285e-06,
      "loss": 0.52,
      "step": 558
    },
    {
      "epoch": 0.99,
      "learning_rate": 4.065907729179149e-06,
      "loss": 0.5167,
      "step": 559
    },
    {
      "epoch": 0.99,
      "learning_rate": 4.062312762133014e-06,
      "loss": 0.4957,
      "step": 560
    },
    {
      "epoch": 1.0,
      "learning_rate": 4.0587177950868785e-06,
      "loss": 0.4721,
      "step": 561
    },
    {
      "epoch": 1.0,
      "learning_rate": 4.055122828040743e-06,
      "loss": 0.5233,
      "step": 562
    },
    {
      "epoch": 1.0,
      "learning_rate": 4.051527860994607e-06,
      "loss": 0.4805,
      "step": 563
    },
    {
      "epoch": 1.0,
      "learning_rate": 4.047932893948472e-06,
      "loss": 0.463,
      "step": 564
    },
    {
      "epoch": 1.0,
      "learning_rate": 4.044337926902337e-06,
      "loss": 0.4476,
      "step": 565
    },
    {
      "epoch": 1.01,
      "learning_rate": 4.040742959856201e-06,
      "loss": 0.4295,
      "step": 566
    },
    {
      "epoch": 1.01,
      "learning_rate": 4.037147992810066e-06,
      "loss": 0.4322,
      "step": 567
    },
    {
      "epoch": 1.01,
      "learning_rate": 4.0335530257639305e-06,
      "loss": 0.4603,
      "step": 568
    },
    {
      "epoch": 1.01,
      "learning_rate": 4.029958058717796e-06,
      "loss": 0.4301,
      "step": 569
    },
    {
      "epoch": 1.01,
      "learning_rate": 4.02636309167166e-06,
      "loss": 0.4357,
      "step": 570
    },
    {
      "epoch": 1.01,
      "learning_rate": 4.022768124625524e-06,
      "loss": 0.4405,
      "step": 571
    },
    {
      "epoch": 1.02,
      "learning_rate": 4.019173157579389e-06,
      "loss": 0.4505,
      "step": 572
    },
    {
      "epoch": 1.02,
      "learning_rate": 4.015578190533254e-06,
      "loss": 0.4326,
      "step": 573
    },
    {
      "epoch": 1.02,
      "learning_rate": 4.011983223487118e-06,
      "loss": 0.4582,
      "step": 574
    },
    {
      "epoch": 1.02,
      "learning_rate": 4.008388256440983e-06,
      "loss": 0.4467,
      "step": 575
    },
    {
      "epoch": 1.02,
      "learning_rate": 4.004793289394847e-06,
      "loss": 0.4213,
      "step": 576
    },
    {
      "epoch": 1.02,
      "learning_rate": 4.001198322348712e-06,
      "loss": 0.4507,
      "step": 577
    },
    {
      "epoch": 1.03,
      "learning_rate": 3.997603355302576e-06,
      "loss": 0.4302,
      "step": 578
    },
    {
      "epoch": 1.03,
      "learning_rate": 3.9940083882564415e-06,
      "loss": 0.4347,
      "step": 579
    },
    {
      "epoch": 1.03,
      "learning_rate": 3.990413421210305e-06,
      "loss": 0.4342,
      "step": 580
    },
    {
      "epoch": 1.03,
      "learning_rate": 3.98681845416417e-06,
      "loss": 0.413,
      "step": 581
    },
    {
      "epoch": 1.03,
      "learning_rate": 3.9832234871180345e-06,
      "loss": 0.4638,
      "step": 582
    },
    {
      "epoch": 1.04,
      "learning_rate": 3.9796285200719e-06,
      "loss": 0.4489,
      "step": 583
    },
    {
      "epoch": 1.04,
      "learning_rate": 3.976033553025764e-06,
      "loss": 0.4497,
      "step": 584
    },
    {
      "epoch": 1.04,
      "learning_rate": 3.972438585979628e-06,
      "loss": 0.4412,
      "step": 585
    },
    {
      "epoch": 1.04,
      "learning_rate": 3.9688436189334935e-06,
      "loss": 0.4346,
      "step": 586
    },
    {
      "epoch": 1.04,
      "learning_rate": 3.965248651887358e-06,
      "loss": 0.4329,
      "step": 587
    },
    {
      "epoch": 1.04,
      "learning_rate": 3.961653684841223e-06,
      "loss": 0.421,
      "step": 588
    },
    {
      "epoch": 1.05,
      "learning_rate": 3.9580587177950864e-06,
      "loss": 0.4415,
      "step": 589
    },
    {
      "epoch": 1.05,
      "learning_rate": 3.954463750748952e-06,
      "loss": 0.4395,
      "step": 590
    },
    {
      "epoch": 1.05,
      "learning_rate": 3.950868783702816e-06,
      "loss": 0.449,
      "step": 591
    },
    {
      "epoch": 1.05,
      "learning_rate": 3.947273816656681e-06,
      "loss": 0.4494,
      "step": 592
    },
    {
      "epoch": 1.05,
      "learning_rate": 3.9436788496105454e-06,
      "loss": 0.4531,
      "step": 593
    },
    {
      "epoch": 1.06,
      "learning_rate": 3.94008388256441e-06,
      "loss": 0.4607,
      "step": 594
    },
    {
      "epoch": 1.06,
      "learning_rate": 3.936488915518274e-06,
      "loss": 0.4379,
      "step": 595
    },
    {
      "epoch": 1.06,
      "learning_rate": 3.932893948472139e-06,
      "loss": 0.4507,
      "step": 596
    },
    {
      "epoch": 1.06,
      "learning_rate": 3.929298981426004e-06,
      "loss": 0.4113,
      "step": 597
    },
    {
      "epoch": 1.06,
      "learning_rate": 3.925704014379869e-06,
      "loss": 0.4306,
      "step": 598
    },
    {
      "epoch": 1.06,
      "learning_rate": 3.922109047333732e-06,
      "loss": 0.4348,
      "step": 599
    },
    {
      "epoch": 1.07,
      "learning_rate": 3.918514080287597e-06,
      "loss": 0.4529,
      "step": 600
    },
    {
      "epoch": 1.07,
      "learning_rate": 3.914919113241462e-06,
      "loss": 0.4225,
      "step": 601
    },
    {
      "epoch": 1.07,
      "learning_rate": 3.911324146195327e-06,
      "loss": 0.4494,
      "step": 602
    },
    {
      "epoch": 1.07,
      "learning_rate": 3.907729179149191e-06,
      "loss": 0.4389,
      "step": 603
    },
    {
      "epoch": 1.07,
      "learning_rate": 3.9041342121030556e-06,
      "loss": 0.4935,
      "step": 604
    },
    {
      "epoch": 1.07,
      "learning_rate": 3.900539245056921e-06,
      "loss": 0.4428,
      "step": 605
    },
    {
      "epoch": 1.08,
      "learning_rate": 3.896944278010785e-06,
      "loss": 0.4121,
      "step": 606
    },
    {
      "epoch": 1.08,
      "learning_rate": 3.89334931096465e-06,
      "loss": 0.4587,
      "step": 607
    },
    {
      "epoch": 1.08,
      "learning_rate": 3.889754343918514e-06,
      "loss": 0.4421,
      "step": 608
    },
    {
      "epoch": 1.08,
      "learning_rate": 3.886159376872379e-06,
      "loss": 0.4326,
      "step": 609
    },
    {
      "epoch": 1.08,
      "learning_rate": 3.882564409826243e-06,
      "loss": 0.4266,
      "step": 610
    },
    {
      "epoch": 1.09,
      "learning_rate": 3.878969442780108e-06,
      "loss": 0.4171,
      "step": 611
    },
    {
      "epoch": 1.09,
      "learning_rate": 3.875374475733973e-06,
      "loss": 0.4161,
      "step": 612
    },
    {
      "epoch": 1.09,
      "learning_rate": 3.871779508687837e-06,
      "loss": 0.4262,
      "step": 613
    },
    {
      "epoch": 1.09,
      "learning_rate": 3.868184541641701e-06,
      "loss": 0.4314,
      "step": 614
    },
    {
      "epoch": 1.09,
      "learning_rate": 3.8645895745955665e-06,
      "loss": 0.4272,
      "step": 615
    },
    {
      "epoch": 1.09,
      "learning_rate": 3.860994607549431e-06,
      "loss": 0.437,
      "step": 616
    },
    {
      "epoch": 1.1,
      "learning_rate": 3.857399640503295e-06,
      "loss": 0.4334,
      "step": 617
    },
    {
      "epoch": 1.1,
      "learning_rate": 3.8538046734571595e-06,
      "loss": 0.4302,
      "step": 618
    },
    {
      "epoch": 1.1,
      "learning_rate": 3.850209706411025e-06,
      "loss": 0.4493,
      "step": 619
    },
    {
      "epoch": 1.1,
      "learning_rate": 3.846614739364889e-06,
      "loss": 0.4416,
      "step": 620
    },
    {
      "epoch": 1.1,
      "learning_rate": 3.843019772318754e-06,
      "loss": 0.4456,
      "step": 621
    },
    {
      "epoch": 1.1,
      "learning_rate": 3.8394248052726185e-06,
      "loss": 0.4357,
      "step": 622
    },
    {
      "epoch": 1.11,
      "learning_rate": 3.835829838226483e-06,
      "loss": 0.4647,
      "step": 623
    },
    {
      "epoch": 1.11,
      "learning_rate": 3.832234871180348e-06,
      "loss": 0.4495,
      "step": 624
    },
    {
      "epoch": 1.11,
      "learning_rate": 3.828639904134212e-06,
      "loss": 0.4355,
      "step": 625
    },
    {
      "epoch": 1.11,
      "learning_rate": 3.8250449370880775e-06,
      "loss": 0.4148,
      "step": 626
    },
    {
      "epoch": 1.11,
      "learning_rate": 3.821449970041941e-06,
      "loss": 0.4451,
      "step": 627
    },
    {
      "epoch": 1.12,
      "learning_rate": 3.817855002995806e-06,
      "loss": 0.4397,
      "step": 628
    },
    {
      "epoch": 1.12,
      "learning_rate": 3.8142600359496705e-06,
      "loss": 0.4597,
      "step": 629
    },
    {
      "epoch": 1.12,
      "learning_rate": 3.8106650689035353e-06,
      "loss": 0.423,
      "step": 630
    },
    {
      "epoch": 1.12,
      "learning_rate": 3.8070701018573996e-06,
      "loss": 0.4804,
      "step": 631
    },
    {
      "epoch": 1.12,
      "learning_rate": 3.8034751348112643e-06,
      "loss": 0.4591,
      "step": 632
    },
    {
      "epoch": 1.12,
      "learning_rate": 3.799880167765129e-06,
      "loss": 0.4405,
      "step": 633
    },
    {
      "epoch": 1.13,
      "learning_rate": 3.796285200718994e-06,
      "loss": 0.4297,
      "step": 634
    },
    {
      "epoch": 1.13,
      "learning_rate": 3.7926902336728586e-06,
      "loss": 0.44,
      "step": 635
    },
    {
      "epoch": 1.13,
      "learning_rate": 3.7890952666267225e-06,
      "loss": 0.4324,
      "step": 636
    },
    {
      "epoch": 1.13,
      "learning_rate": 3.7855002995805872e-06,
      "loss": 0.4386,
      "step": 637
    },
    {
      "epoch": 1.13,
      "learning_rate": 3.781905332534452e-06,
      "loss": 0.4356,
      "step": 638
    },
    {
      "epoch": 1.13,
      "learning_rate": 3.7783103654883167e-06,
      "loss": 0.4316,
      "step": 639
    },
    {
      "epoch": 1.14,
      "learning_rate": 3.7747153984421806e-06,
      "loss": 0.4333,
      "step": 640
    },
    {
      "epoch": 1.14,
      "learning_rate": 3.7711204313960454e-06,
      "loss": 0.4479,
      "step": 641
    },
    {
      "epoch": 1.14,
      "learning_rate": 3.76752546434991e-06,
      "loss": 0.4613,
      "step": 642
    },
    {
      "epoch": 1.14,
      "learning_rate": 3.763930497303775e-06,
      "loss": 0.4532,
      "step": 643
    },
    {
      "epoch": 1.14,
      "learning_rate": 3.7603355302576396e-06,
      "loss": 0.4495,
      "step": 644
    },
    {
      "epoch": 1.15,
      "learning_rate": 3.7567405632115035e-06,
      "loss": 0.4344,
      "step": 645
    },
    {
      "epoch": 1.15,
      "learning_rate": 3.7531455961653683e-06,
      "loss": 0.4476,
      "step": 646
    },
    {
      "epoch": 1.15,
      "learning_rate": 3.749550629119233e-06,
      "loss": 0.4236,
      "step": 647
    },
    {
      "epoch": 1.15,
      "learning_rate": 3.7459556620730978e-06,
      "loss": 0.4529,
      "step": 648
    },
    {
      "epoch": 1.15,
      "learning_rate": 3.7423606950269625e-06,
      "loss": 0.4271,
      "step": 649
    },
    {
      "epoch": 1.15,
      "learning_rate": 3.738765727980827e-06,
      "loss": 0.4217,
      "step": 650
    },
    {
      "epoch": 1.16,
      "learning_rate": 3.7351707609346916e-06,
      "loss": 0.4328,
      "step": 651
    },
    {
      "epoch": 1.16,
      "learning_rate": 3.7315757938885564e-06,
      "loss": 0.4205,
      "step": 652
    },
    {
      "epoch": 1.16,
      "learning_rate": 3.727980826842421e-06,
      "loss": 0.4308,
      "step": 653
    },
    {
      "epoch": 1.16,
      "learning_rate": 3.724385859796285e-06,
      "loss": 0.4433,
      "step": 654
    },
    {
      "epoch": 1.16,
      "learning_rate": 3.7207908927501498e-06,
      "loss": 0.4369,
      "step": 655
    },
    {
      "epoch": 1.17,
      "learning_rate": 3.7171959257040145e-06,
      "loss": 0.4207,
      "step": 656
    },
    {
      "epoch": 1.17,
      "learning_rate": 3.7136009586578793e-06,
      "loss": 0.4384,
      "step": 657
    },
    {
      "epoch": 1.17,
      "learning_rate": 3.710005991611744e-06,
      "loss": 0.4743,
      "step": 658
    },
    {
      "epoch": 1.17,
      "learning_rate": 3.706411024565608e-06,
      "loss": 0.4644,
      "step": 659
    },
    {
      "epoch": 1.17,
      "learning_rate": 3.7028160575194727e-06,
      "loss": 0.4378,
      "step": 660
    },
    {
      "epoch": 1.17,
      "learning_rate": 3.6992210904733374e-06,
      "loss": 0.4325,
      "step": 661
    },
    {
      "epoch": 1.18,
      "learning_rate": 3.695626123427202e-06,
      "loss": 0.4443,
      "step": 662
    },
    {
      "epoch": 1.18,
      "learning_rate": 3.692031156381067e-06,
      "loss": 0.4364,
      "step": 663
    },
    {
      "epoch": 1.18,
      "learning_rate": 3.688436189334931e-06,
      "loss": 0.4329,
      "step": 664
    },
    {
      "epoch": 1.18,
      "learning_rate": 3.6848412222887956e-06,
      "loss": 0.453,
      "step": 665
    },
    {
      "epoch": 1.18,
      "learning_rate": 3.6812462552426603e-06,
      "loss": 0.448,
      "step": 666
    },
    {
      "epoch": 1.18,
      "learning_rate": 3.677651288196525e-06,
      "loss": 0.4244,
      "step": 667
    },
    {
      "epoch": 1.19,
      "learning_rate": 3.6740563211503894e-06,
      "loss": 0.4537,
      "step": 668
    },
    {
      "epoch": 1.19,
      "learning_rate": 3.670461354104254e-06,
      "loss": 0.4306,
      "step": 669
    },
    {
      "epoch": 1.19,
      "learning_rate": 3.666866387058119e-06,
      "loss": 0.428,
      "step": 670
    },
    {
      "epoch": 1.19,
      "learning_rate": 3.6632714200119836e-06,
      "loss": 0.454,
      "step": 671
    },
    {
      "epoch": 1.19,
      "learning_rate": 3.6596764529658484e-06,
      "loss": 0.3971,
      "step": 672
    },
    {
      "epoch": 1.2,
      "learning_rate": 3.6560814859197123e-06,
      "loss": 0.4513,
      "step": 673
    },
    {
      "epoch": 1.2,
      "learning_rate": 3.652486518873577e-06,
      "loss": 0.4284,
      "step": 674
    },
    {
      "epoch": 1.2,
      "learning_rate": 3.6488915518274418e-06,
      "loss": 0.4211,
      "step": 675
    },
    {
      "epoch": 1.2,
      "learning_rate": 3.6452965847813065e-06,
      "loss": 0.4482,
      "step": 676
    },
    {
      "epoch": 1.2,
      "learning_rate": 3.6417016177351713e-06,
      "loss": 0.4549,
      "step": 677
    },
    {
      "epoch": 1.2,
      "learning_rate": 3.638106650689035e-06,
      "loss": 0.4402,
      "step": 678
    },
    {
      "epoch": 1.21,
      "learning_rate": 3.6345116836429e-06,
      "loss": 0.423,
      "step": 679
    },
    {
      "epoch": 1.21,
      "learning_rate": 3.6309167165967647e-06,
      "loss": 0.42,
      "step": 680
    },
    {
      "epoch": 1.21,
      "learning_rate": 3.6273217495506294e-06,
      "loss": 0.4471,
      "step": 681
    },
    {
      "epoch": 1.21,
      "learning_rate": 3.6237267825044933e-06,
      "loss": 0.4687,
      "step": 682
    },
    {
      "epoch": 1.21,
      "learning_rate": 3.620131815458358e-06,
      "loss": 0.451,
      "step": 683
    },
    {
      "epoch": 1.21,
      "learning_rate": 3.616536848412223e-06,
      "loss": 0.4157,
      "step": 684
    },
    {
      "epoch": 1.22,
      "learning_rate": 3.6129418813660876e-06,
      "loss": 0.4449,
      "step": 685
    },
    {
      "epoch": 1.22,
      "learning_rate": 3.6093469143199523e-06,
      "loss": 0.445,
      "step": 686
    },
    {
      "epoch": 1.22,
      "learning_rate": 3.6057519472738167e-06,
      "loss": 0.4523,
      "step": 687
    },
    {
      "epoch": 1.22,
      "learning_rate": 3.6021569802276814e-06,
      "loss": 0.462,
      "step": 688
    },
    {
      "epoch": 1.22,
      "learning_rate": 3.598562013181546e-06,
      "loss": 0.4373,
      "step": 689
    },
    {
      "epoch": 1.23,
      "learning_rate": 3.594967046135411e-06,
      "loss": 0.4382,
      "step": 690
    },
    {
      "epoch": 1.23,
      "learning_rate": 3.591372079089275e-06,
      "loss": 0.4482,
      "step": 691
    },
    {
      "epoch": 1.23,
      "learning_rate": 3.5877771120431396e-06,
      "loss": 0.4609,
      "step": 692
    },
    {
      "epoch": 1.23,
      "learning_rate": 3.5841821449970043e-06,
      "loss": 0.419,
      "step": 693
    },
    {
      "epoch": 1.23,
      "learning_rate": 3.580587177950869e-06,
      "loss": 0.4535,
      "step": 694
    },
    {
      "epoch": 1.23,
      "learning_rate": 3.576992210904734e-06,
      "loss": 0.4328,
      "step": 695
    },
    {
      "epoch": 1.24,
      "learning_rate": 3.5733972438585977e-06,
      "loss": 0.4232,
      "step": 696
    },
    {
      "epoch": 1.24,
      "learning_rate": 3.5698022768124625e-06,
      "loss": 0.4596,
      "step": 697
    },
    {
      "epoch": 1.24,
      "learning_rate": 3.5662073097663272e-06,
      "loss": 0.4169,
      "step": 698
    },
    {
      "epoch": 1.24,
      "learning_rate": 3.562612342720192e-06,
      "loss": 0.4692,
      "step": 699
    },
    {
      "epoch": 1.24,
      "learning_rate": 3.5590173756740567e-06,
      "loss": 0.4453,
      "step": 700
    },
    {
      "epoch": 1.25,
      "learning_rate": 3.5554224086279206e-06,
      "loss": 0.4391,
      "step": 701
    },
    {
      "epoch": 1.25,
      "learning_rate": 3.5518274415817854e-06,
      "loss": 0.4285,
      "step": 702
    },
    {
      "epoch": 1.25,
      "learning_rate": 3.54823247453565e-06,
      "loss": 0.4266,
      "step": 703
    },
    {
      "epoch": 1.25,
      "learning_rate": 3.544637507489515e-06,
      "loss": 0.4496,
      "step": 704
    },
    {
      "epoch": 1.25,
      "learning_rate": 3.541042540443379e-06,
      "loss": 0.452,
      "step": 705
    },
    {
      "epoch": 1.25,
      "learning_rate": 3.537447573397244e-06,
      "loss": 0.4218,
      "step": 706
    },
    {
      "epoch": 1.26,
      "learning_rate": 3.5338526063511087e-06,
      "loss": 0.4009,
      "step": 707
    },
    {
      "epoch": 1.26,
      "learning_rate": 3.5302576393049734e-06,
      "loss": 0.418,
      "step": 708
    },
    {
      "epoch": 1.26,
      "learning_rate": 3.526662672258838e-06,
      "loss": 0.4651,
      "step": 709
    },
    {
      "epoch": 1.26,
      "learning_rate": 3.523067705212702e-06,
      "loss": 0.4366,
      "step": 710
    },
    {
      "epoch": 1.26,
      "learning_rate": 3.519472738166567e-06,
      "loss": 0.4584,
      "step": 711
    },
    {
      "epoch": 1.26,
      "learning_rate": 3.5158777711204316e-06,
      "loss": 0.4282,
      "step": 712
    },
    {
      "epoch": 1.27,
      "learning_rate": 3.5122828040742963e-06,
      "loss": 0.433,
      "step": 713
    },
    {
      "epoch": 1.27,
      "learning_rate": 3.508687837028161e-06,
      "loss": 0.4204,
      "step": 714
    },
    {
      "epoch": 1.27,
      "learning_rate": 3.505092869982025e-06,
      "loss": 0.4501,
      "step": 715
    },
    {
      "epoch": 1.27,
      "learning_rate": 3.5014979029358897e-06,
      "loss": 0.4307,
      "step": 716
    },
    {
      "epoch": 1.27,
      "learning_rate": 3.4979029358897545e-06,
      "loss": 0.4593,
      "step": 717
    },
    {
      "epoch": 1.28,
      "learning_rate": 3.4943079688436192e-06,
      "loss": 0.4526,
      "step": 718
    },
    {
      "epoch": 1.28,
      "learning_rate": 3.490713001797483e-06,
      "loss": 0.4635,
      "step": 719
    },
    {
      "epoch": 1.28,
      "learning_rate": 3.487118034751348e-06,
      "loss": 0.4358,
      "step": 720
    },
    {
      "epoch": 1.28,
      "learning_rate": 3.4835230677052126e-06,
      "loss": 0.4252,
      "step": 721
    },
    {
      "epoch": 1.28,
      "learning_rate": 3.4799281006590774e-06,
      "loss": 0.4103,
      "step": 722
    },
    {
      "epoch": 1.28,
      "learning_rate": 3.476333133612942e-06,
      "loss": 0.4201,
      "step": 723
    },
    {
      "epoch": 1.29,
      "learning_rate": 3.4727381665668065e-06,
      "loss": 0.4323,
      "step": 724
    },
    {
      "epoch": 1.29,
      "learning_rate": 3.4691431995206712e-06,
      "loss": 0.4063,
      "step": 725
    },
    {
      "epoch": 1.29,
      "learning_rate": 3.465548232474536e-06,
      "loss": 0.4259,
      "step": 726
    },
    {
      "epoch": 1.29,
      "learning_rate": 3.4619532654284007e-06,
      "loss": 0.4419,
      "step": 727
    },
    {
      "epoch": 1.29,
      "learning_rate": 3.458358298382265e-06,
      "loss": 0.4255,
      "step": 728
    },
    {
      "epoch": 1.29,
      "learning_rate": 3.4547633313361294e-06,
      "loss": 0.4533,
      "step": 729
    },
    {
      "epoch": 1.3,
      "learning_rate": 3.451168364289994e-06,
      "loss": 0.4193,
      "step": 730
    },
    {
      "epoch": 1.3,
      "learning_rate": 3.447573397243859e-06,
      "loss": 0.4448,
      "step": 731
    },
    {
      "epoch": 1.3,
      "learning_rate": 3.4439784301977236e-06,
      "loss": 0.4458,
      "step": 732
    },
    {
      "epoch": 1.3,
      "learning_rate": 3.4403834631515875e-06,
      "loss": 0.4514,
      "step": 733
    },
    {
      "epoch": 1.3,
      "learning_rate": 3.4367884961054523e-06,
      "loss": 0.4044,
      "step": 734
    },
    {
      "epoch": 1.31,
      "learning_rate": 3.433193529059317e-06,
      "loss": 0.4463,
      "step": 735
    },
    {
      "epoch": 1.31,
      "learning_rate": 3.4295985620131818e-06,
      "loss": 0.465,
      "step": 736
    },
    {
      "epoch": 1.31,
      "learning_rate": 3.4260035949670465e-06,
      "loss": 0.4374,
      "step": 737
    },
    {
      "epoch": 1.31,
      "learning_rate": 3.4224086279209104e-06,
      "loss": 0.4212,
      "step": 738
    },
    {
      "epoch": 1.31,
      "learning_rate": 3.418813660874775e-06,
      "loss": 0.4151,
      "step": 739
    },
    {
      "epoch": 1.31,
      "learning_rate": 3.41521869382864e-06,
      "loss": 0.4382,
      "step": 740
    },
    {
      "epoch": 1.32,
      "learning_rate": 3.4116237267825047e-06,
      "loss": 0.4348,
      "step": 741
    },
    {
      "epoch": 1.32,
      "learning_rate": 3.408028759736369e-06,
      "loss": 0.4324,
      "step": 742
    },
    {
      "epoch": 1.32,
      "learning_rate": 3.4044337926902337e-06,
      "loss": 0.4362,
      "step": 743
    },
    {
      "epoch": 1.32,
      "learning_rate": 3.4008388256440985e-06,
      "loss": 0.4633,
      "step": 744
    },
    {
      "epoch": 1.32,
      "learning_rate": 3.3972438585979632e-06,
      "loss": 0.4246,
      "step": 745
    },
    {
      "epoch": 1.33,
      "learning_rate": 3.3936488915518276e-06,
      "loss": 0.4244,
      "step": 746
    },
    {
      "epoch": 1.33,
      "learning_rate": 3.390053924505692e-06,
      "loss": 0.4398,
      "step": 747
    },
    {
      "epoch": 1.33,
      "learning_rate": 3.3864589574595566e-06,
      "loss": 0.4388,
      "step": 748
    },
    {
      "epoch": 1.33,
      "learning_rate": 3.3828639904134214e-06,
      "loss": 0.4394,
      "step": 749
    },
    {
      "epoch": 1.33,
      "learning_rate": 3.379269023367286e-06,
      "loss": 0.4569,
      "step": 750
    },
    {
      "epoch": 1.33,
      "learning_rate": 3.375674056321151e-06,
      "loss": 0.4484,
      "step": 751
    },
    {
      "epoch": 1.34,
      "learning_rate": 3.372079089275015e-06,
      "loss": 0.4724,
      "step": 752
    },
    {
      "epoch": 1.34,
      "learning_rate": 3.3684841222288795e-06,
      "loss": 0.4403,
      "step": 753
    },
    {
      "epoch": 1.34,
      "learning_rate": 3.3648891551827443e-06,
      "loss": 0.4307,
      "step": 754
    },
    {
      "epoch": 1.34,
      "learning_rate": 3.361294188136609e-06,
      "loss": 0.4482,
      "step": 755
    },
    {
      "epoch": 1.34,
      "learning_rate": 3.357699221090473e-06,
      "loss": 0.425,
      "step": 756
    },
    {
      "epoch": 1.34,
      "learning_rate": 3.3541042540443377e-06,
      "loss": 0.4546,
      "step": 757
    },
    {
      "epoch": 1.35,
      "learning_rate": 3.3505092869982024e-06,
      "loss": 0.4571,
      "step": 758
    },
    {
      "epoch": 1.35,
      "learning_rate": 3.346914319952067e-06,
      "loss": 0.4193,
      "step": 759
    },
    {
      "epoch": 1.35,
      "learning_rate": 3.343319352905932e-06,
      "loss": 0.4275,
      "step": 760
    },
    {
      "epoch": 1.35,
      "learning_rate": 3.3397243858597963e-06,
      "loss": 0.415,
      "step": 761
    },
    {
      "epoch": 1.35,
      "learning_rate": 3.336129418813661e-06,
      "loss": 0.4439,
      "step": 762
    },
    {
      "epoch": 1.36,
      "learning_rate": 3.3325344517675258e-06,
      "loss": 0.4265,
      "step": 763
    },
    {
      "epoch": 1.36,
      "learning_rate": 3.3289394847213905e-06,
      "loss": 0.4312,
      "step": 764
    },
    {
      "epoch": 1.36,
      "learning_rate": 3.325344517675255e-06,
      "loss": 0.4605,
      "step": 765
    },
    {
      "epoch": 1.36,
      "learning_rate": 3.321749550629119e-06,
      "loss": 0.414,
      "step": 766
    },
    {
      "epoch": 1.36,
      "learning_rate": 3.318154583582984e-06,
      "loss": 0.421,
      "step": 767
    },
    {
      "epoch": 1.36,
      "learning_rate": 3.3145596165368487e-06,
      "loss": 0.4203,
      "step": 768
    },
    {
      "epoch": 1.37,
      "learning_rate": 3.3109646494907134e-06,
      "loss": 0.4166,
      "step": 769
    },
    {
      "epoch": 1.37,
      "learning_rate": 3.3073696824445773e-06,
      "loss": 0.4008,
      "step": 770
    },
    {
      "epoch": 1.37,
      "learning_rate": 3.303774715398442e-06,
      "loss": 0.4218,
      "step": 771
    },
    {
      "epoch": 1.37,
      "learning_rate": 3.300179748352307e-06,
      "loss": 0.4511,
      "step": 772
    },
    {
      "epoch": 1.37,
      "learning_rate": 3.2965847813061716e-06,
      "loss": 0.4303,
      "step": 773
    },
    {
      "epoch": 1.37,
      "learning_rate": 3.2929898142600363e-06,
      "loss": 0.4198,
      "step": 774
    },
    {
      "epoch": 1.38,
      "learning_rate": 3.2893948472139002e-06,
      "loss": 0.4332,
      "step": 775
    },
    {
      "epoch": 1.38,
      "learning_rate": 3.285799880167765e-06,
      "loss": 0.4274,
      "step": 776
    },
    {
      "epoch": 1.38,
      "learning_rate": 3.2822049131216297e-06,
      "loss": 0.4326,
      "step": 777
    },
    {
      "epoch": 1.38,
      "learning_rate": 3.2786099460754945e-06,
      "loss": 0.447,
      "step": 778
    },
    {
      "epoch": 1.38,
      "learning_rate": 3.2750149790293592e-06,
      "loss": 0.4371,
      "step": 779
    },
    {
      "epoch": 1.39,
      "learning_rate": 3.2714200119832236e-06,
      "loss": 0.4341,
      "step": 780
    },
    {
      "epoch": 1.39,
      "learning_rate": 3.2678250449370883e-06,
      "loss": 0.4243,
      "step": 781
    },
    {
      "epoch": 1.39,
      "learning_rate": 3.264230077890953e-06,
      "loss": 0.4411,
      "step": 782
    },
    {
      "epoch": 1.39,
      "learning_rate": 3.2606351108448174e-06,
      "loss": 0.4238,
      "step": 783
    },
    {
      "epoch": 1.39,
      "learning_rate": 3.2570401437986817e-06,
      "loss": 0.4431,
      "step": 784
    },
    {
      "epoch": 1.39,
      "learning_rate": 3.2534451767525465e-06,
      "loss": 0.4546,
      "step": 785
    },
    {
      "epoch": 1.4,
      "learning_rate": 3.249850209706411e-06,
      "loss": 0.4407,
      "step": 786
    },
    {
      "epoch": 1.4,
      "learning_rate": 3.246255242660276e-06,
      "loss": 0.4449,
      "step": 787
    },
    {
      "epoch": 1.4,
      "learning_rate": 3.2426602756141407e-06,
      "loss": 0.4329,
      "step": 788
    },
    {
      "epoch": 1.4,
      "learning_rate": 3.2390653085680046e-06,
      "loss": 0.4413,
      "step": 789
    },
    {
      "epoch": 1.4,
      "learning_rate": 3.2354703415218694e-06,
      "loss": 0.4098,
      "step": 790
    },
    {
      "epoch": 1.4,
      "learning_rate": 3.231875374475734e-06,
      "loss": 0.4288,
      "step": 791
    },
    {
      "epoch": 1.41,
      "learning_rate": 3.228280407429599e-06,
      "loss": 0.4207,
      "step": 792
    },
    {
      "epoch": 1.41,
      "learning_rate": 3.2246854403834628e-06,
      "loss": 0.4335,
      "step": 793
    },
    {
      "epoch": 1.41,
      "learning_rate": 3.2210904733373275e-06,
      "loss": 0.445,
      "step": 794
    },
    {
      "epoch": 1.41,
      "learning_rate": 3.2174955062911923e-06,
      "loss": 0.4138,
      "step": 795
    },
    {
      "epoch": 1.41,
      "learning_rate": 3.213900539245057e-06,
      "loss": 0.4626,
      "step": 796
    },
    {
      "epoch": 1.42,
      "learning_rate": 3.2103055721989218e-06,
      "loss": 0.4419,
      "step": 797
    },
    {
      "epoch": 1.42,
      "learning_rate": 3.206710605152786e-06,
      "loss": 0.4493,
      "step": 798
    },
    {
      "epoch": 1.42,
      "learning_rate": 3.203115638106651e-06,
      "loss": 0.4337,
      "step": 799
    },
    {
      "epoch": 1.42,
      "learning_rate": 3.1995206710605156e-06,
      "loss": 0.446,
      "step": 800
    },
    {
      "epoch": 1.42,
      "learning_rate": 3.19592570401438e-06,
      "loss": 0.4176,
      "step": 801
    },
    {
      "epoch": 1.42,
      "learning_rate": 3.1923307369682447e-06,
      "loss": 0.4635,
      "step": 802
    },
    {
      "epoch": 1.43,
      "learning_rate": 3.188735769922109e-06,
      "loss": 0.4326,
      "step": 803
    },
    {
      "epoch": 1.43,
      "learning_rate": 3.1851408028759737e-06,
      "loss": 0.4454,
      "step": 804
    },
    {
      "epoch": 1.43,
      "learning_rate": 3.1815458358298385e-06,
      "loss": 0.4368,
      "step": 805
    },
    {
      "epoch": 1.43,
      "learning_rate": 3.1779508687837032e-06,
      "loss": 0.4485,
      "step": 806
    },
    {
      "epoch": 1.43,
      "learning_rate": 3.174355901737567e-06,
      "loss": 0.4172,
      "step": 807
    },
    {
      "epoch": 1.44,
      "learning_rate": 3.170760934691432e-06,
      "loss": 0.4466,
      "step": 808
    },
    {
      "epoch": 1.44,
      "learning_rate": 3.1671659676452966e-06,
      "loss": 0.4371,
      "step": 809
    },
    {
      "epoch": 1.44,
      "learning_rate": 3.1635710005991614e-06,
      "loss": 0.4421,
      "step": 810
    },
    {
      "epoch": 1.44,
      "learning_rate": 3.159976033553026e-06,
      "loss": 0.4436,
      "step": 811
    },
    {
      "epoch": 1.44,
      "learning_rate": 3.15638106650689e-06,
      "loss": 0.434,
      "step": 812
    },
    {
      "epoch": 1.44,
      "learning_rate": 3.1527860994607548e-06,
      "loss": 0.4344,
      "step": 813
    },
    {
      "epoch": 1.45,
      "learning_rate": 3.1491911324146195e-06,
      "loss": 0.4567,
      "step": 814
    },
    {
      "epoch": 1.45,
      "learning_rate": 3.1455961653684843e-06,
      "loss": 0.422,
      "step": 815
    },
    {
      "epoch": 1.45,
      "learning_rate": 3.142001198322349e-06,
      "loss": 0.4215,
      "step": 816
    },
    {
      "epoch": 1.45,
      "learning_rate": 3.1384062312762134e-06,
      "loss": 0.4335,
      "step": 817
    },
    {
      "epoch": 1.45,
      "learning_rate": 3.134811264230078e-06,
      "loss": 0.4697,
      "step": 818
    },
    {
      "epoch": 1.45,
      "learning_rate": 3.1312162971839424e-06,
      "loss": 0.4409,
      "step": 819
    },
    {
      "epoch": 1.46,
      "learning_rate": 3.127621330137807e-06,
      "loss": 0.4363,
      "step": 820
    },
    {
      "epoch": 1.46,
      "learning_rate": 3.1240263630916715e-06,
      "loss": 0.4324,
      "step": 821
    },
    {
      "epoch": 1.46,
      "learning_rate": 3.1204313960455363e-06,
      "loss": 0.4359,
      "step": 822
    },
    {
      "epoch": 1.46,
      "learning_rate": 3.116836428999401e-06,
      "loss": 0.4419,
      "step": 823
    },
    {
      "epoch": 1.46,
      "learning_rate": 3.1132414619532658e-06,
      "loss": 0.4174,
      "step": 824
    },
    {
      "epoch": 1.47,
      "learning_rate": 3.1096464949071305e-06,
      "loss": 0.4297,
      "step": 825
    },
    {
      "epoch": 1.47,
      "learning_rate": 3.1060515278609944e-06,
      "loss": 0.4319,
      "step": 826
    },
    {
      "epoch": 1.47,
      "learning_rate": 3.102456560814859e-06,
      "loss": 0.4445,
      "step": 827
    },
    {
      "epoch": 1.47,
      "learning_rate": 3.098861593768724e-06,
      "loss": 0.4538,
      "step": 828
    },
    {
      "epoch": 1.47,
      "learning_rate": 3.0952666267225887e-06,
      "loss": 0.4253,
      "step": 829
    },
    {
      "epoch": 1.47,
      "learning_rate": 3.0916716596764534e-06,
      "loss": 0.4406,
      "step": 830
    },
    {
      "epoch": 1.48,
      "learning_rate": 3.0880766926303173e-06,
      "loss": 0.4385,
      "step": 831
    },
    {
      "epoch": 1.48,
      "learning_rate": 3.084481725584182e-06,
      "loss": 0.4495,
      "step": 832
    },
    {
      "epoch": 1.48,
      "learning_rate": 3.080886758538047e-06,
      "loss": 0.4377,
      "step": 833
    },
    {
      "epoch": 1.48,
      "learning_rate": 3.0772917914919116e-06,
      "loss": 0.4066,
      "step": 834
    },
    {
      "epoch": 1.48,
      "learning_rate": 3.073696824445776e-06,
      "loss": 0.4284,
      "step": 835
    },
    {
      "epoch": 1.48,
      "learning_rate": 3.0701018573996406e-06,
      "loss": 0.4379,
      "step": 836
    },
    {
      "epoch": 1.49,
      "learning_rate": 3.0665068903535054e-06,
      "loss": 0.4316,
      "step": 837
    },
    {
      "epoch": 1.49,
      "learning_rate": 3.0629119233073697e-06,
      "loss": 0.4492,
      "step": 838
    },
    {
      "epoch": 1.49,
      "learning_rate": 3.0593169562612345e-06,
      "loss": 0.4297,
      "step": 839
    },
    {
      "epoch": 1.49,
      "learning_rate": 3.0557219892150988e-06,
      "loss": 0.4677,
      "step": 840
    },
    {
      "epoch": 1.49,
      "learning_rate": 3.0521270221689635e-06,
      "loss": 0.4503,
      "step": 841
    },
    {
      "epoch": 1.5,
      "learning_rate": 3.0485320551228283e-06,
      "loss": 0.429,
      "step": 842
    },
    {
      "epoch": 1.5,
      "learning_rate": 3.044937088076693e-06,
      "loss": 0.4376,
      "step": 843
    },
    {
      "epoch": 1.5,
      "learning_rate": 3.041342121030557e-06,
      "loss": 0.4188,
      "step": 844
    },
    {
      "epoch": 1.5,
      "learning_rate": 3.0377471539844217e-06,
      "loss": 0.4184,
      "step": 845
    },
    {
      "epoch": 1.5,
      "learning_rate": 3.0341521869382864e-06,
      "loss": 0.4492,
      "step": 846
    },
    {
      "epoch": 1.5,
      "learning_rate": 3.030557219892151e-06,
      "loss": 0.4393,
      "step": 847
    },
    {
      "epoch": 1.51,
      "learning_rate": 3.026962252846016e-06,
      "loss": 0.4182,
      "step": 848
    },
    {
      "epoch": 1.51,
      "learning_rate": 3.02336728579988e-06,
      "loss": 0.4378,
      "step": 849
    },
    {
      "epoch": 1.51,
      "learning_rate": 3.0197723187537446e-06,
      "loss": 0.4379,
      "step": 850
    },
    {
      "epoch": 1.51,
      "learning_rate": 3.0161773517076093e-06,
      "loss": 0.4315,
      "step": 851
    },
    {
      "epoch": 1.51,
      "learning_rate": 3.012582384661474e-06,
      "loss": 0.4263,
      "step": 852
    },
    {
      "epoch": 1.52,
      "learning_rate": 3.008987417615339e-06,
      "loss": 0.436,
      "step": 853
    },
    {
      "epoch": 1.52,
      "learning_rate": 3.005392450569203e-06,
      "loss": 0.4307,
      "step": 854
    },
    {
      "epoch": 1.52,
      "learning_rate": 3.001797483523068e-06,
      "loss": 0.4567,
      "step": 855
    },
    {
      "epoch": 1.52,
      "learning_rate": 2.9982025164769322e-06,
      "loss": 0.447,
      "step": 856
    },
    {
      "epoch": 1.52,
      "learning_rate": 2.994607549430797e-06,
      "loss": 0.4342,
      "step": 857
    },
    {
      "epoch": 1.52,
      "learning_rate": 2.9910125823846617e-06,
      "loss": 0.4591,
      "step": 858
    },
    {
      "epoch": 1.53,
      "learning_rate": 2.987417615338526e-06,
      "loss": 0.4358,
      "step": 859
    },
    {
      "epoch": 1.53,
      "learning_rate": 2.983822648292391e-06,
      "loss": 0.4298,
      "step": 860
    },
    {
      "epoch": 1.53,
      "learning_rate": 2.9802276812462556e-06,
      "loss": 0.4525,
      "step": 861
    },
    {
      "epoch": 1.53,
      "learning_rate": 2.97663271420012e-06,
      "loss": 0.444,
      "step": 862
    },
    {
      "epoch": 1.53,
      "learning_rate": 2.9730377471539846e-06,
      "loss": 0.4298,
      "step": 863
    },
    {
      "epoch": 1.53,
      "learning_rate": 2.969442780107849e-06,
      "loss": 0.4185,
      "step": 864
    },
    {
      "epoch": 1.54,
      "learning_rate": 2.9658478130617137e-06,
      "loss": 0.4573,
      "step": 865
    },
    {
      "epoch": 1.54,
      "learning_rate": 2.962252846015578e-06,
      "loss": 0.4381,
      "step": 866
    },
    {
      "epoch": 1.54,
      "learning_rate": 2.958657878969443e-06,
      "loss": 0.4286,
      "step": 867
    },
    {
      "epoch": 1.54,
      "learning_rate": 2.9550629119233075e-06,
      "loss": 0.4482,
      "step": 868
    },
    {
      "epoch": 1.54,
      "learning_rate": 2.951467944877172e-06,
      "loss": 0.4016,
      "step": 869
    },
    {
      "epoch": 1.55,
      "learning_rate": 2.9478729778310366e-06,
      "loss": 0.4517,
      "step": 870
    },
    {
      "epoch": 1.55,
      "learning_rate": 2.944278010784901e-06,
      "loss": 0.4178,
      "step": 871
    },
    {
      "epoch": 1.55,
      "learning_rate": 2.9406830437387657e-06,
      "loss": 0.4494,
      "step": 872
    },
    {
      "epoch": 1.55,
      "learning_rate": 2.9370880766926304e-06,
      "loss": 0.4177,
      "step": 873
    },
    {
      "epoch": 1.55,
      "learning_rate": 2.9334931096464948e-06,
      "loss": 0.4359,
      "step": 874
    },
    {
      "epoch": 1.55,
      "learning_rate": 2.9298981426003595e-06,
      "loss": 0.4301,
      "step": 875
    },
    {
      "epoch": 1.56,
      "learning_rate": 2.9263031755542243e-06,
      "loss": 0.4461,
      "step": 876
    },
    {
      "epoch": 1.56,
      "learning_rate": 2.922708208508089e-06,
      "loss": 0.4204,
      "step": 877
    },
    {
      "epoch": 1.56,
      "learning_rate": 2.9191132414619533e-06,
      "loss": 0.4363,
      "step": 878
    },
    {
      "epoch": 1.56,
      "learning_rate": 2.915518274415818e-06,
      "loss": 0.4376,
      "step": 879
    },
    {
      "epoch": 1.56,
      "learning_rate": 2.9119233073696824e-06,
      "loss": 0.441,
      "step": 880
    },
    {
      "epoch": 1.56,
      "learning_rate": 2.908328340323547e-06,
      "loss": 0.4361,
      "step": 881
    },
    {
      "epoch": 1.57,
      "learning_rate": 2.904733373277412e-06,
      "loss": 0.4508,
      "step": 882
    },
    {
      "epoch": 1.57,
      "learning_rate": 2.9011384062312762e-06,
      "loss": 0.4315,
      "step": 883
    },
    {
      "epoch": 1.57,
      "learning_rate": 2.897543439185141e-06,
      "loss": 0.4249,
      "step": 884
    },
    {
      "epoch": 1.57,
      "learning_rate": 2.8939484721390053e-06,
      "loss": 0.4602,
      "step": 885
    },
    {
      "epoch": 1.57,
      "learning_rate": 2.89035350509287e-06,
      "loss": 0.4386,
      "step": 886
    },
    {
      "epoch": 1.58,
      "learning_rate": 2.8867585380467344e-06,
      "loss": 0.466,
      "step": 887
    },
    {
      "epoch": 1.58,
      "learning_rate": 2.883163571000599e-06,
      "loss": 0.4574,
      "step": 888
    },
    {
      "epoch": 1.58,
      "learning_rate": 2.879568603954464e-06,
      "loss": 0.4115,
      "step": 889
    },
    {
      "epoch": 1.58,
      "learning_rate": 2.8759736369083282e-06,
      "loss": 0.4395,
      "step": 890
    },
    {
      "epoch": 1.58,
      "learning_rate": 2.872378669862193e-06,
      "loss": 0.4206,
      "step": 891
    },
    {
      "epoch": 1.58,
      "learning_rate": 2.8687837028160577e-06,
      "loss": 0.4274,
      "step": 892
    },
    {
      "epoch": 1.59,
      "learning_rate": 2.865188735769922e-06,
      "loss": 0.4431,
      "step": 893
    },
    {
      "epoch": 1.59,
      "learning_rate": 2.861593768723787e-06,
      "loss": 0.4285,
      "step": 894
    },
    {
      "epoch": 1.59,
      "learning_rate": 2.8579988016776515e-06,
      "loss": 0.4259,
      "step": 895
    },
    {
      "epoch": 1.59,
      "learning_rate": 2.8544038346315163e-06,
      "loss": 0.4314,
      "step": 896
    },
    {
      "epoch": 1.59,
      "learning_rate": 2.8508088675853806e-06,
      "loss": 0.4092,
      "step": 897
    },
    {
      "epoch": 1.6,
      "learning_rate": 2.8472139005392454e-06,
      "loss": 0.3943,
      "step": 898
    },
    {
      "epoch": 1.6,
      "learning_rate": 2.8436189334931097e-06,
      "loss": 0.4184,
      "step": 899
    },
    {
      "epoch": 1.6,
      "learning_rate": 2.8400239664469744e-06,
      "loss": 0.4348,
      "step": 900
    },
    {
      "epoch": 1.6,
      "learning_rate": 2.8364289994008388e-06,
      "loss": 0.4588,
      "step": 901
    },
    {
      "epoch": 1.6,
      "learning_rate": 2.8328340323547035e-06,
      "loss": 0.4242,
      "step": 902
    },
    {
      "epoch": 1.6,
      "learning_rate": 2.8292390653085683e-06,
      "loss": 0.4279,
      "step": 903
    },
    {
      "epoch": 1.61,
      "learning_rate": 2.8256440982624326e-06,
      "loss": 0.4616,
      "step": 904
    },
    {
      "epoch": 1.61,
      "learning_rate": 2.8220491312162973e-06,
      "loss": 0.4153,
      "step": 905
    },
    {
      "epoch": 1.61,
      "learning_rate": 2.8184541641701617e-06,
      "loss": 0.4445,
      "step": 906
    },
    {
      "epoch": 1.61,
      "learning_rate": 2.8148591971240264e-06,
      "loss": 0.4062,
      "step": 907
    },
    {
      "epoch": 1.61,
      "learning_rate": 2.8112642300778907e-06,
      "loss": 0.4313,
      "step": 908
    },
    {
      "epoch": 1.61,
      "learning_rate": 2.8076692630317555e-06,
      "loss": 0.4519,
      "step": 909
    },
    {
      "epoch": 1.62,
      "learning_rate": 2.8040742959856202e-06,
      "loss": 0.4149,
      "step": 910
    },
    {
      "epoch": 1.62,
      "learning_rate": 2.8004793289394846e-06,
      "loss": 0.4223,
      "step": 911
    },
    {
      "epoch": 1.62,
      "learning_rate": 2.7968843618933493e-06,
      "loss": 0.4504,
      "step": 912
    },
    {
      "epoch": 1.62,
      "learning_rate": 2.793289394847214e-06,
      "loss": 0.4193,
      "step": 913
    },
    {
      "epoch": 1.62,
      "learning_rate": 2.789694427801079e-06,
      "loss": 0.4321,
      "step": 914
    },
    {
      "epoch": 1.63,
      "learning_rate": 2.786099460754943e-06,
      "loss": 0.411,
      "step": 915
    },
    {
      "epoch": 1.63,
      "learning_rate": 2.782504493708808e-06,
      "loss": 0.4197,
      "step": 916
    },
    {
      "epoch": 1.63,
      "learning_rate": 2.7789095266626722e-06,
      "loss": 0.4369,
      "step": 917
    },
    {
      "epoch": 1.63,
      "learning_rate": 2.775314559616537e-06,
      "loss": 0.4394,
      "step": 918
    },
    {
      "epoch": 1.63,
      "learning_rate": 2.7717195925704017e-06,
      "loss": 0.4422,
      "step": 919
    },
    {
      "epoch": 1.63,
      "learning_rate": 2.768124625524266e-06,
      "loss": 0.4452,
      "step": 920
    },
    {
      "epoch": 1.64,
      "learning_rate": 2.764529658478131e-06,
      "loss": 0.404,
      "step": 921
    },
    {
      "epoch": 1.64,
      "learning_rate": 2.760934691431995e-06,
      "loss": 0.4237,
      "step": 922
    },
    {
      "epoch": 1.64,
      "learning_rate": 2.75733972438586e-06,
      "loss": 0.4315,
      "step": 923
    },
    {
      "epoch": 1.64,
      "learning_rate": 2.753744757339724e-06,
      "loss": 0.4305,
      "step": 924
    },
    {
      "epoch": 1.64,
      "learning_rate": 2.750149790293589e-06,
      "loss": 0.4167,
      "step": 925
    },
    {
      "epoch": 1.64,
      "learning_rate": 2.7465548232474537e-06,
      "loss": 0.4142,
      "step": 926
    },
    {
      "epoch": 1.65,
      "learning_rate": 2.742959856201318e-06,
      "loss": 0.4341,
      "step": 927
    },
    {
      "epoch": 1.65,
      "learning_rate": 2.7393648891551828e-06,
      "loss": 0.4471,
      "step": 928
    },
    {
      "epoch": 1.65,
      "learning_rate": 2.735769922109047e-06,
      "loss": 0.4426,
      "step": 929
    },
    {
      "epoch": 1.65,
      "learning_rate": 2.732174955062912e-06,
      "loss": 0.42,
      "step": 930
    },
    {
      "epoch": 1.65,
      "learning_rate": 2.7285799880167766e-06,
      "loss": 0.4405,
      "step": 931
    },
    {
      "epoch": 1.66,
      "learning_rate": 2.7249850209706414e-06,
      "loss": 0.458,
      "step": 932
    },
    {
      "epoch": 1.66,
      "learning_rate": 2.721390053924506e-06,
      "loss": 0.4027,
      "step": 933
    },
    {
      "epoch": 1.66,
      "learning_rate": 2.7177950868783704e-06,
      "loss": 0.4239,
      "step": 934
    },
    {
      "epoch": 1.66,
      "learning_rate": 2.714200119832235e-06,
      "loss": 0.4504,
      "step": 935
    },
    {
      "epoch": 1.66,
      "learning_rate": 2.7106051527860995e-06,
      "loss": 0.4296,
      "step": 936
    },
    {
      "epoch": 1.66,
      "learning_rate": 2.7070101857399643e-06,
      "loss": 0.4171,
      "step": 937
    },
    {
      "epoch": 1.67,
      "learning_rate": 2.7034152186938286e-06,
      "loss": 0.4219,
      "step": 938
    },
    {
      "epoch": 1.67,
      "learning_rate": 2.6998202516476933e-06,
      "loss": 0.4244,
      "step": 939
    },
    {
      "epoch": 1.67,
      "learning_rate": 2.696225284601558e-06,
      "loss": 0.4472,
      "step": 940
    },
    {
      "epoch": 1.67,
      "learning_rate": 2.6926303175554224e-06,
      "loss": 0.4226,
      "step": 941
    },
    {
      "epoch": 1.67,
      "learning_rate": 2.689035350509287e-06,
      "loss": 0.4333,
      "step": 942
    },
    {
      "epoch": 1.67,
      "learning_rate": 2.6854403834631515e-06,
      "loss": 0.4545,
      "step": 943
    },
    {
      "epoch": 1.68,
      "learning_rate": 2.6818454164170162e-06,
      "loss": 0.4249,
      "step": 944
    },
    {
      "epoch": 1.68,
      "learning_rate": 2.6782504493708806e-06,
      "loss": 0.419,
      "step": 945
    },
    {
      "epoch": 1.68,
      "learning_rate": 2.6746554823247453e-06,
      "loss": 0.445,
      "step": 946
    },
    {
      "epoch": 1.68,
      "learning_rate": 2.67106051527861e-06,
      "loss": 0.4191,
      "step": 947
    },
    {
      "epoch": 1.68,
      "learning_rate": 2.6674655482324744e-06,
      "loss": 0.4176,
      "step": 948
    },
    {
      "epoch": 1.69,
      "learning_rate": 2.663870581186339e-06,
      "loss": 0.4372,
      "step": 949
    },
    {
      "epoch": 1.69,
      "learning_rate": 2.660275614140204e-06,
      "loss": 0.4208,
      "step": 950
    },
    {
      "epoch": 1.69,
      "learning_rate": 2.6566806470940686e-06,
      "loss": 0.4406,
      "step": 951
    },
    {
      "epoch": 1.69,
      "learning_rate": 2.653085680047933e-06,
      "loss": 0.4349,
      "step": 952
    },
    {
      "epoch": 1.69,
      "learning_rate": 2.6494907130017977e-06,
      "loss": 0.4418,
      "step": 953
    },
    {
      "epoch": 1.69,
      "learning_rate": 2.6458957459556625e-06,
      "loss": 0.4566,
      "step": 954
    },
    {
      "epoch": 1.7,
      "learning_rate": 2.6423007789095268e-06,
      "loss": 0.4342,
      "step": 955
    },
    {
      "epoch": 1.7,
      "learning_rate": 2.6387058118633915e-06,
      "loss": 0.4462,
      "step": 956
    },
    {
      "epoch": 1.7,
      "learning_rate": 2.635110844817256e-06,
      "loss": 0.4175,
      "step": 957
    },
    {
      "epoch": 1.7,
      "learning_rate": 2.6315158777711206e-06,
      "loss": 0.4361,
      "step": 958
    },
    {
      "epoch": 1.7,
      "learning_rate": 2.627920910724985e-06,
      "loss": 0.4373,
      "step": 959
    },
    {
      "epoch": 1.71,
      "learning_rate": 2.6243259436788497e-06,
      "loss": 0.4544,
      "step": 960
    },
    {
      "epoch": 1.71,
      "learning_rate": 2.620730976632714e-06,
      "loss": 0.4169,
      "step": 961
    },
    {
      "epoch": 1.71,
      "learning_rate": 2.6171360095865788e-06,
      "loss": 0.4206,
      "step": 962
    },
    {
      "epoch": 1.71,
      "learning_rate": 2.6135410425404435e-06,
      "loss": 0.4599,
      "step": 963
    },
    {
      "epoch": 1.71,
      "learning_rate": 2.609946075494308e-06,
      "loss": 0.4569,
      "step": 964
    },
    {
      "epoch": 1.71,
      "learning_rate": 2.6063511084481726e-06,
      "loss": 0.4592,
      "step": 965
    },
    {
      "epoch": 1.72,
      "learning_rate": 2.602756141402037e-06,
      "loss": 0.4607,
      "step": 966
    },
    {
      "epoch": 1.72,
      "learning_rate": 2.5991611743559017e-06,
      "loss": 0.4314,
      "step": 967
    },
    {
      "epoch": 1.72,
      "learning_rate": 2.5955662073097664e-06,
      "loss": 0.4433,
      "step": 968
    },
    {
      "epoch": 1.72,
      "learning_rate": 2.591971240263631e-06,
      "loss": 0.4599,
      "step": 969
    },
    {
      "epoch": 1.72,
      "learning_rate": 2.588376273217496e-06,
      "loss": 0.4108,
      "step": 970
    },
    {
      "epoch": 1.72,
      "learning_rate": 2.5847813061713602e-06,
      "loss": 0.4203,
      "step": 971
    },
    {
      "epoch": 1.73,
      "learning_rate": 2.581186339125225e-06,
      "loss": 0.425,
      "step": 972
    },
    {
      "epoch": 1.73,
      "learning_rate": 2.5775913720790893e-06,
      "loss": 0.4184,
      "step": 973
    },
    {
      "epoch": 1.73,
      "learning_rate": 2.573996405032954e-06,
      "loss": 0.4594,
      "step": 974
    },
    {
      "epoch": 1.73,
      "learning_rate": 2.5704014379868184e-06,
      "loss": 0.4294,
      "step": 975
    },
    {
      "epoch": 1.73,
      "learning_rate": 2.566806470940683e-06,
      "loss": 0.45,
      "step": 976
    },
    {
      "epoch": 1.74,
      "learning_rate": 2.563211503894548e-06,
      "loss": 0.4245,
      "step": 977
    },
    {
      "epoch": 1.74,
      "learning_rate": 2.559616536848412e-06,
      "loss": 0.4216,
      "step": 978
    },
    {
      "epoch": 1.74,
      "learning_rate": 2.556021569802277e-06,
      "loss": 0.4354,
      "step": 979
    },
    {
      "epoch": 1.74,
      "learning_rate": 2.5524266027561413e-06,
      "loss": 0.4369,
      "step": 980
    },
    {
      "epoch": 1.74,
      "learning_rate": 2.548831635710006e-06,
      "loss": 0.4141,
      "step": 981
    },
    {
      "epoch": 1.74,
      "learning_rate": 2.5452366686638704e-06,
      "loss": 0.4199,
      "step": 982
    },
    {
      "epoch": 1.75,
      "learning_rate": 2.541641701617735e-06,
      "loss": 0.4488,
      "step": 983
    },
    {
      "epoch": 1.75,
      "learning_rate": 2.5380467345716e-06,
      "loss": 0.4315,
      "step": 984
    },
    {
      "epoch": 1.75,
      "learning_rate": 2.534451767525464e-06,
      "loss": 0.4111,
      "step": 985
    },
    {
      "epoch": 1.75,
      "learning_rate": 2.530856800479329e-06,
      "loss": 0.4549,
      "step": 986
    },
    {
      "epoch": 1.75,
      "learning_rate": 2.5272618334331937e-06,
      "loss": 0.4448,
      "step": 987
    },
    {
      "epoch": 1.75,
      "learning_rate": 2.5236668663870584e-06,
      "loss": 0.4353,
      "step": 988
    },
    {
      "epoch": 1.76,
      "learning_rate": 2.5200718993409228e-06,
      "loss": 0.4429,
      "step": 989
    },
    {
      "epoch": 1.76,
      "learning_rate": 2.5164769322947875e-06,
      "loss": 0.4385,
      "step": 990
    },
    {
      "epoch": 1.76,
      "learning_rate": 2.5128819652486523e-06,
      "loss": 0.4174,
      "step": 991
    },
    {
      "epoch": 1.76,
      "learning_rate": 2.5092869982025166e-06,
      "loss": 0.4346,
      "step": 992
    },
    {
      "epoch": 1.76,
      "learning_rate": 2.5056920311563813e-06,
      "loss": 0.4402,
      "step": 993
    },
    {
      "epoch": 1.77,
      "learning_rate": 2.5020970641102457e-06,
      "loss": 0.4221,
      "step": 994
    },
    {
      "epoch": 1.77,
      "learning_rate": 2.4985020970641104e-06,
      "loss": 0.4262,
      "step": 995
    },
    {
      "epoch": 1.77,
      "learning_rate": 2.4949071300179747e-06,
      "loss": 0.4569,
      "step": 996
    },
    {
      "epoch": 1.77,
      "learning_rate": 2.4913121629718395e-06,
      "loss": 0.4329,
      "step": 997
    },
    {
      "epoch": 1.77,
      "learning_rate": 2.4877171959257042e-06,
      "loss": 0.4469,
      "step": 998
    },
    {
      "epoch": 1.77,
      "learning_rate": 2.4841222288795686e-06,
      "loss": 0.4293,
      "step": 999
    },
    {
      "epoch": 1.78,
      "learning_rate": 2.4805272618334333e-06,
      "loss": 0.4149,
      "step": 1000
    },
    {
      "epoch": 1.78,
      "learning_rate": 2.4769322947872976e-06,
      "loss": 0.4442,
      "step": 1001
    },
    {
      "epoch": 1.78,
      "learning_rate": 2.4733373277411624e-06,
      "loss": 0.4226,
      "step": 1002
    },
    {
      "epoch": 1.78,
      "learning_rate": 2.4697423606950267e-06,
      "loss": 0.4367,
      "step": 1003
    },
    {
      "epoch": 1.78,
      "learning_rate": 2.4661473936488915e-06,
      "loss": 0.4437,
      "step": 1004
    },
    {
      "epoch": 1.79,
      "learning_rate": 2.4625524266027562e-06,
      "loss": 0.4152,
      "step": 1005
    },
    {
      "epoch": 1.79,
      "learning_rate": 2.458957459556621e-06,
      "loss": 0.4091,
      "step": 1006
    },
    {
      "epoch": 1.79,
      "learning_rate": 2.4553624925104857e-06,
      "loss": 0.4392,
      "step": 1007
    },
    {
      "epoch": 1.79,
      "learning_rate": 2.45176752546435e-06,
      "loss": 0.41,
      "step": 1008
    },
    {
      "epoch": 1.79,
      "learning_rate": 2.4481725584182148e-06,
      "loss": 0.4304,
      "step": 1009
    },
    {
      "epoch": 1.79,
      "learning_rate": 2.444577591372079e-06,
      "loss": 0.4131,
      "step": 1010
    },
    {
      "epoch": 1.8,
      "learning_rate": 2.440982624325944e-06,
      "loss": 0.4437,
      "step": 1011
    },
    {
      "epoch": 1.8,
      "learning_rate": 2.437387657279808e-06,
      "loss": 0.4396,
      "step": 1012
    },
    {
      "epoch": 1.8,
      "learning_rate": 2.433792690233673e-06,
      "loss": 0.4476,
      "step": 1013
    },
    {
      "epoch": 1.8,
      "learning_rate": 2.4301977231875377e-06,
      "loss": 0.4501,
      "step": 1014
    },
    {
      "epoch": 1.8,
      "learning_rate": 2.426602756141402e-06,
      "loss": 0.4013,
      "step": 1015
    },
    {
      "epoch": 1.8,
      "learning_rate": 2.4230077890952668e-06,
      "loss": 0.4307,
      "step": 1016
    },
    {
      "epoch": 1.81,
      "learning_rate": 2.419412822049131e-06,
      "loss": 0.4385,
      "step": 1017
    },
    {
      "epoch": 1.81,
      "learning_rate": 2.415817855002996e-06,
      "loss": 0.4619,
      "step": 1018
    },
    {
      "epoch": 1.81,
      "learning_rate": 2.41222288795686e-06,
      "loss": 0.4358,
      "step": 1019
    },
    {
      "epoch": 1.81,
      "learning_rate": 2.408627920910725e-06,
      "loss": 0.415,
      "step": 1020
    },
    {
      "epoch": 1.81,
      "learning_rate": 2.4050329538645897e-06,
      "loss": 0.4417,
      "step": 1021
    },
    {
      "epoch": 1.82,
      "learning_rate": 2.401437986818454e-06,
      "loss": 0.4156,
      "step": 1022
    },
    {
      "epoch": 1.82,
      "learning_rate": 2.3978430197723187e-06,
      "loss": 0.4565,
      "step": 1023
    },
    {
      "epoch": 1.82,
      "learning_rate": 2.3942480527261835e-06,
      "loss": 0.4655,
      "step": 1024
    },
    {
      "epoch": 1.82,
      "learning_rate": 2.3906530856800482e-06,
      "loss": 0.4228,
      "step": 1025
    },
    {
      "epoch": 1.82,
      "learning_rate": 2.3870581186339126e-06,
      "loss": 0.4586,
      "step": 1026
    },
    {
      "epoch": 1.82,
      "learning_rate": 2.3834631515877773e-06,
      "loss": 0.4115,
      "step": 1027
    },
    {
      "epoch": 1.83,
      "learning_rate": 2.379868184541642e-06,
      "loss": 0.4084,
      "step": 1028
    },
    {
      "epoch": 1.83,
      "learning_rate": 2.3762732174955064e-06,
      "loss": 0.4154,
      "step": 1029
    },
    {
      "epoch": 1.83,
      "learning_rate": 2.372678250449371e-06,
      "loss": 0.4508,
      "step": 1030
    },
    {
      "epoch": 1.83,
      "learning_rate": 2.3690832834032355e-06,
      "loss": 0.4086,
      "step": 1031
    },
    {
      "epoch": 1.83,
      "learning_rate": 2.3654883163571002e-06,
      "loss": 0.4345,
      "step": 1032
    },
    {
      "epoch": 1.83,
      "learning_rate": 2.3618933493109645e-06,
      "loss": 0.4329,
      "step": 1033
    },
    {
      "epoch": 1.84,
      "learning_rate": 2.3582983822648293e-06,
      "loss": 0.4297,
      "step": 1034
    },
    {
      "epoch": 1.84,
      "learning_rate": 2.354703415218694e-06,
      "loss": 0.4206,
      "step": 1035
    },
    {
      "epoch": 1.84,
      "learning_rate": 2.3511084481725584e-06,
      "loss": 0.4353,
      "step": 1036
    },
    {
      "epoch": 1.84,
      "learning_rate": 2.347513481126423e-06,
      "loss": 0.4364,
      "step": 1037
    },
    {
      "epoch": 1.84,
      "learning_rate": 2.3439185140802874e-06,
      "loss": 0.4508,
      "step": 1038
    },
    {
      "epoch": 1.85,
      "learning_rate": 2.340323547034152e-06,
      "loss": 0.4358,
      "step": 1039
    },
    {
      "epoch": 1.85,
      "learning_rate": 2.3367285799880165e-06,
      "loss": 0.4328,
      "step": 1040
    },
    {
      "epoch": 1.85,
      "learning_rate": 2.3331336129418813e-06,
      "loss": 0.4037,
      "step": 1041
    },
    {
      "epoch": 1.85,
      "learning_rate": 2.329538645895746e-06,
      "loss": 0.4546,
      "step": 1042
    },
    {
      "epoch": 1.85,
      "learning_rate": 2.3259436788496108e-06,
      "loss": 0.4081,
      "step": 1043
    },
    {
      "epoch": 1.85,
      "learning_rate": 2.322348711803475e-06,
      "loss": 0.4143,
      "step": 1044
    },
    {
      "epoch": 1.86,
      "learning_rate": 2.31875374475734e-06,
      "loss": 0.4628,
      "step": 1045
    },
    {
      "epoch": 1.86,
      "learning_rate": 2.3151587777112046e-06,
      "loss": 0.4386,
      "step": 1046
    },
    {
      "epoch": 1.86,
      "learning_rate": 2.311563810665069e-06,
      "loss": 0.447,
      "step": 1047
    },
    {
      "epoch": 1.86,
      "learning_rate": 2.3079688436189337e-06,
      "loss": 0.4417,
      "step": 1048
    },
    {
      "epoch": 1.86,
      "learning_rate": 2.3043738765727984e-06,
      "loss": 0.4447,
      "step": 1049
    },
    {
      "epoch": 1.87,
      "learning_rate": 2.3007789095266627e-06,
      "loss": 0.4128,
      "step": 1050
    },
    {
      "epoch": 1.87,
      "learning_rate": 2.2971839424805275e-06,
      "loss": 0.4185,
      "step": 1051
    },
    {
      "epoch": 1.87,
      "learning_rate": 2.293588975434392e-06,
      "loss": 0.4128,
      "step": 1052
    },
    {
      "epoch": 1.87,
      "learning_rate": 2.2899940083882566e-06,
      "loss": 0.4246,
      "step": 1053
    },
    {
      "epoch": 1.87,
      "learning_rate": 2.286399041342121e-06,
      "loss": 0.4238,
      "step": 1054
    },
    {
      "epoch": 1.87,
      "learning_rate": 2.2828040742959856e-06,
      "loss": 0.4529,
      "step": 1055
    },
    {
      "epoch": 1.88,
      "learning_rate": 2.2792091072498504e-06,
      "loss": 0.4176,
      "step": 1056
    },
    {
      "epoch": 1.88,
      "learning_rate": 2.2756141402037147e-06,
      "loss": 0.4357,
      "step": 1057
    },
    {
      "epoch": 1.88,
      "learning_rate": 2.2720191731575795e-06,
      "loss": 0.4264,
      "step": 1058
    },
    {
      "epoch": 1.88,
      "learning_rate": 2.268424206111444e-06,
      "loss": 0.4423,
      "step": 1059
    },
    {
      "epoch": 1.88,
      "learning_rate": 2.2648292390653085e-06,
      "loss": 0.4301,
      "step": 1060
    },
    {
      "epoch": 1.88,
      "learning_rate": 2.2612342720191733e-06,
      "loss": 0.4354,
      "step": 1061
    },
    {
      "epoch": 1.89,
      "learning_rate": 2.2576393049730376e-06,
      "loss": 0.413,
      "step": 1062
    },
    {
      "epoch": 1.89,
      "learning_rate": 2.2540443379269024e-06,
      "loss": 0.4537,
      "step": 1063
    },
    {
      "epoch": 1.89,
      "learning_rate": 2.250449370880767e-06,
      "loss": 0.4369,
      "step": 1064
    },
    {
      "epoch": 1.89,
      "learning_rate": 2.246854403834632e-06,
      "loss": 0.4405,
      "step": 1065
    },
    {
      "epoch": 1.89,
      "learning_rate": 2.243259436788496e-06,
      "loss": 0.4033,
      "step": 1066
    },
    {
      "epoch": 1.9,
      "learning_rate": 2.239664469742361e-06,
      "loss": 0.4321,
      "step": 1067
    },
    {
      "epoch": 1.9,
      "learning_rate": 2.2360695026962253e-06,
      "loss": 0.438,
      "step": 1068
    },
    {
      "epoch": 1.9,
      "learning_rate": 2.23247453565009e-06,
      "loss": 0.4152,
      "step": 1069
    },
    {
      "epoch": 1.9,
      "learning_rate": 2.2288795686039543e-06,
      "loss": 0.4133,
      "step": 1070
    },
    {
      "epoch": 1.9,
      "learning_rate": 2.225284601557819e-06,
      "loss": 0.4164,
      "step": 1071
    },
    {
      "epoch": 1.9,
      "learning_rate": 2.221689634511684e-06,
      "loss": 0.4213,
      "step": 1072
    },
    {
      "epoch": 1.91,
      "learning_rate": 2.218094667465548e-06,
      "loss": 0.448,
      "step": 1073
    },
    {
      "epoch": 1.91,
      "learning_rate": 2.214499700419413e-06,
      "loss": 0.4263,
      "step": 1074
    },
    {
      "epoch": 1.91,
      "learning_rate": 2.2109047333732773e-06,
      "loss": 0.4185,
      "step": 1075
    },
    {
      "epoch": 1.91,
      "learning_rate": 2.207309766327142e-06,
      "loss": 0.4319,
      "step": 1076
    },
    {
      "epoch": 1.91,
      "learning_rate": 2.2037147992810063e-06,
      "loss": 0.4258,
      "step": 1077
    },
    {
      "epoch": 1.91,
      "learning_rate": 2.200119832234871e-06,
      "loss": 0.4294,
      "step": 1078
    },
    {
      "epoch": 1.92,
      "learning_rate": 2.196524865188736e-06,
      "loss": 0.4465,
      "step": 1079
    },
    {
      "epoch": 1.92,
      "learning_rate": 2.1929298981426006e-06,
      "loss": 0.4403,
      "step": 1080
    },
    {
      "epoch": 1.92,
      "learning_rate": 2.189334931096465e-06,
      "loss": 0.4191,
      "step": 1081
    },
    {
      "epoch": 1.92,
      "learning_rate": 2.1857399640503297e-06,
      "loss": 0.4182,
      "step": 1082
    },
    {
      "epoch": 1.92,
      "learning_rate": 2.1821449970041944e-06,
      "loss": 0.4339,
      "step": 1083
    },
    {
      "epoch": 1.93,
      "learning_rate": 2.1785500299580587e-06,
      "loss": 0.4275,
      "step": 1084
    },
    {
      "epoch": 1.93,
      "learning_rate": 2.1749550629119235e-06,
      "loss": 0.4383,
      "step": 1085
    },
    {
      "epoch": 1.93,
      "learning_rate": 2.1713600958657882e-06,
      "loss": 0.4496,
      "step": 1086
    },
    {
      "epoch": 1.93,
      "learning_rate": 2.1677651288196526e-06,
      "loss": 0.432,
      "step": 1087
    },
    {
      "epoch": 1.93,
      "learning_rate": 2.1641701617735173e-06,
      "loss": 0.4521,
      "step": 1088
    },
    {
      "epoch": 1.93,
      "learning_rate": 2.1605751947273816e-06,
      "loss": 0.4338,
      "step": 1089
    },
    {
      "epoch": 1.94,
      "learning_rate": 2.1569802276812464e-06,
      "loss": 0.4299,
      "step": 1090
    },
    {
      "epoch": 1.94,
      "learning_rate": 2.1533852606351107e-06,
      "loss": 0.4585,
      "step": 1091
    },
    {
      "epoch": 1.94,
      "learning_rate": 2.1497902935889755e-06,
      "loss": 0.4616,
      "step": 1092
    },
    {
      "epoch": 1.94,
      "learning_rate": 2.14619532654284e-06,
      "loss": 0.4336,
      "step": 1093
    },
    {
      "epoch": 1.94,
      "learning_rate": 2.1426003594967045e-06,
      "loss": 0.4517,
      "step": 1094
    },
    {
      "epoch": 1.94,
      "learning_rate": 2.1390053924505693e-06,
      "loss": 0.4471,
      "step": 1095
    },
    {
      "epoch": 1.95,
      "learning_rate": 2.1354104254044336e-06,
      "loss": 0.4218,
      "step": 1096
    },
    {
      "epoch": 1.95,
      "learning_rate": 2.1318154583582984e-06,
      "loss": 0.4151,
      "step": 1097
    },
    {
      "epoch": 1.95,
      "learning_rate": 2.128220491312163e-06,
      "loss": 0.4289,
      "step": 1098
    },
    {
      "epoch": 1.95,
      "learning_rate": 2.1246255242660274e-06,
      "loss": 0.4509,
      "step": 1099
    },
    {
      "epoch": 1.95,
      "learning_rate": 2.121030557219892e-06,
      "loss": 0.4298,
      "step": 1100
    },
    {
      "epoch": 1.96,
      "learning_rate": 2.117435590173757e-06,
      "loss": 0.4184,
      "step": 1101
    },
    {
      "epoch": 1.96,
      "learning_rate": 2.1138406231276217e-06,
      "loss": 0.4283,
      "step": 1102
    },
    {
      "epoch": 1.96,
      "learning_rate": 2.110245656081486e-06,
      "loss": 0.4343,
      "step": 1103
    },
    {
      "epoch": 1.96,
      "learning_rate": 2.1066506890353508e-06,
      "loss": 0.4377,
      "step": 1104
    },
    {
      "epoch": 1.96,
      "learning_rate": 2.103055721989215e-06,
      "loss": 0.4462,
      "step": 1105
    },
    {
      "epoch": 1.96,
      "learning_rate": 2.09946075494308e-06,
      "loss": 0.4375,
      "step": 1106
    },
    {
      "epoch": 1.97,
      "learning_rate": 2.0958657878969446e-06,
      "loss": 0.4504,
      "step": 1107
    },
    {
      "epoch": 1.97,
      "learning_rate": 2.092270820850809e-06,
      "loss": 0.4319,
      "step": 1108
    },
    {
      "epoch": 1.97,
      "learning_rate": 2.0886758538046737e-06,
      "loss": 0.4208,
      "step": 1109
    },
    {
      "epoch": 1.97,
      "learning_rate": 2.085080886758538e-06,
      "loss": 0.4281,
      "step": 1110
    },
    {
      "epoch": 1.97,
      "learning_rate": 2.0814859197124027e-06,
      "loss": 0.4326,
      "step": 1111
    },
    {
      "epoch": 1.98,
      "learning_rate": 2.077890952666267e-06,
      "loss": 0.4224,
      "step": 1112
    },
    {
      "epoch": 1.98,
      "learning_rate": 2.074295985620132e-06,
      "loss": 0.4078,
      "step": 1113
    },
    {
      "epoch": 1.98,
      "learning_rate": 2.070701018573996e-06,
      "loss": 0.4374,
      "step": 1114
    },
    {
      "epoch": 1.98,
      "learning_rate": 2.067106051527861e-06,
      "loss": 0.4485,
      "step": 1115
    },
    {
      "epoch": 1.98,
      "learning_rate": 2.0635110844817256e-06,
      "loss": 0.4208,
      "step": 1116
    },
    {
      "epoch": 1.98,
      "learning_rate": 2.05991611743559e-06,
      "loss": 0.4208,
      "step": 1117
    },
    {
      "epoch": 1.99,
      "learning_rate": 2.0563211503894547e-06,
      "loss": 0.457,
      "step": 1118
    },
    {
      "epoch": 1.99,
      "learning_rate": 2.0527261833433195e-06,
      "loss": 0.4436,
      "step": 1119
    },
    {
      "epoch": 1.99,
      "learning_rate": 2.049131216297184e-06,
      "loss": 0.4491,
      "step": 1120
    },
    {
      "epoch": 1.99,
      "learning_rate": 2.0455362492510485e-06,
      "loss": 0.4123,
      "step": 1121
    },
    {
      "epoch": 1.99,
      "learning_rate": 2.0419412822049133e-06,
      "loss": 0.4235,
      "step": 1122
    },
    {
      "epoch": 1.99,
      "learning_rate": 2.038346315158778e-06,
      "loss": 0.444,
      "step": 1123
    },
    {
      "epoch": 2.0,
      "learning_rate": 2.0347513481126424e-06,
      "loss": 0.4422,
      "step": 1124
    },
    {
      "epoch": 2.0,
      "learning_rate": 2.031156381066507e-06,
      "loss": 0.4411,
      "step": 1125
    },
    {
      "epoch": 2.0,
      "learning_rate": 2.0275614140203714e-06,
      "loss": 0.407,
      "step": 1126
    },
    {
      "epoch": 2.0,
      "learning_rate": 2.023966446974236e-06,
      "loss": 0.4073,
      "step": 1127
    },
    {
      "epoch": 2.0,
      "learning_rate": 2.0203714799281005e-06,
      "loss": 0.3811,
      "step": 1128
    },
    {
      "epoch": 2.01,
      "learning_rate": 2.0167765128819653e-06,
      "loss": 0.383,
      "step": 1129
    },
    {
      "epoch": 2.01,
      "learning_rate": 2.01318154583583e-06,
      "loss": 0.3999,
      "step": 1130
    },
    {
      "epoch": 2.01,
      "learning_rate": 2.0095865787896943e-06,
      "loss": 0.3781,
      "step": 1131
    },
    {
      "epoch": 2.01,
      "learning_rate": 2.005991611743559e-06,
      "loss": 0.3927,
      "step": 1132
    },
    {
      "epoch": 2.01,
      "learning_rate": 2.0023966446974234e-06,
      "loss": 0.3926,
      "step": 1133
    },
    {
      "epoch": 2.01,
      "learning_rate": 1.998801677651288e-06,
      "loss": 0.3843,
      "step": 1134
    },
    {
      "epoch": 2.02,
      "learning_rate": 1.9952067106051525e-06,
      "loss": 0.3792,
      "step": 1135
    },
    {
      "epoch": 2.02,
      "learning_rate": 1.9916117435590172e-06,
      "loss": 0.3766,
      "step": 1136
    },
    {
      "epoch": 2.02,
      "learning_rate": 1.988016776512882e-06,
      "loss": 0.3713,
      "step": 1137
    },
    {
      "epoch": 2.02,
      "learning_rate": 1.9844218094667467e-06,
      "loss": 0.4266,
      "step": 1138
    },
    {
      "epoch": 2.02,
      "learning_rate": 1.9808268424206115e-06,
      "loss": 0.3931,
      "step": 1139
    },
    {
      "epoch": 2.02,
      "learning_rate": 1.977231875374476e-06,
      "loss": 0.3851,
      "step": 1140
    },
    {
      "epoch": 2.03,
      "learning_rate": 1.9736369083283406e-06,
      "loss": 0.3972,
      "step": 1141
    },
    {
      "epoch": 2.03,
      "learning_rate": 1.970041941282205e-06,
      "loss": 0.3478,
      "step": 1142
    },
    {
      "epoch": 2.03,
      "learning_rate": 1.9664469742360696e-06,
      "loss": 0.3962,
      "step": 1143
    },
    {
      "epoch": 2.03,
      "learning_rate": 1.9628520071899344e-06,
      "loss": 0.4133,
      "step": 1144
    },
    {
      "epoch": 2.03,
      "learning_rate": 1.9592570401437987e-06,
      "loss": 0.3934,
      "step": 1145
    },
    {
      "epoch": 2.04,
      "learning_rate": 1.9556620730976635e-06,
      "loss": 0.3785,
      "step": 1146
    },
    {
      "epoch": 2.04,
      "learning_rate": 1.9520671060515278e-06,
      "loss": 0.395,
      "step": 1147
    },
    {
      "epoch": 2.04,
      "learning_rate": 1.9484721390053925e-06,
      "loss": 0.4045,
      "step": 1148
    },
    {
      "epoch": 2.04,
      "learning_rate": 1.944877171959257e-06,
      "loss": 0.3676,
      "step": 1149
    },
    {
      "epoch": 2.04,
      "learning_rate": 1.9412822049131216e-06,
      "loss": 0.3877,
      "step": 1150
    },
    {
      "epoch": 2.04,
      "learning_rate": 1.9376872378669864e-06,
      "loss": 0.3615,
      "step": 1151
    },
    {
      "epoch": 2.05,
      "learning_rate": 1.9340922708208507e-06,
      "loss": 0.4009,
      "step": 1152
    },
    {
      "epoch": 2.05,
      "learning_rate": 1.9304973037747154e-06,
      "loss": 0.3842,
      "step": 1153
    },
    {
      "epoch": 2.05,
      "learning_rate": 1.9269023367285798e-06,
      "loss": 0.3672,
      "step": 1154
    },
    {
      "epoch": 2.05,
      "learning_rate": 1.9233073696824445e-06,
      "loss": 0.3904,
      "step": 1155
    },
    {
      "epoch": 2.05,
      "learning_rate": 1.9197124026363093e-06,
      "loss": 0.3921,
      "step": 1156
    },
    {
      "epoch": 2.06,
      "learning_rate": 1.916117435590174e-06,
      "loss": 0.3778,
      "step": 1157
    },
    {
      "epoch": 2.06,
      "learning_rate": 1.9125224685440388e-06,
      "loss": 0.3769,
      "step": 1158
    },
    {
      "epoch": 2.06,
      "learning_rate": 1.908927501497903e-06,
      "loss": 0.3934,
      "step": 1159
    },
    {
      "epoch": 2.06,
      "learning_rate": 1.9053325344517676e-06,
      "loss": 0.3738,
      "step": 1160
    },
    {
      "epoch": 2.06,
      "learning_rate": 1.9017375674056322e-06,
      "loss": 0.3788,
      "step": 1161
    },
    {
      "epoch": 2.06,
      "learning_rate": 1.898142600359497e-06,
      "loss": 0.412,
      "step": 1162
    },
    {
      "epoch": 2.07,
      "learning_rate": 1.8945476333133612e-06,
      "loss": 0.3828,
      "step": 1163
    },
    {
      "epoch": 2.07,
      "learning_rate": 1.890952666267226e-06,
      "loss": 0.3634,
      "step": 1164
    },
    {
      "epoch": 2.07,
      "learning_rate": 1.8873576992210903e-06,
      "loss": 0.3984,
      "step": 1165
    },
    {
      "epoch": 2.07,
      "learning_rate": 1.883762732174955e-06,
      "loss": 0.3851,
      "step": 1166
    },
    {
      "epoch": 2.07,
      "learning_rate": 1.8801677651288198e-06,
      "loss": 0.4053,
      "step": 1167
    },
    {
      "epoch": 2.07,
      "learning_rate": 1.8765727980826841e-06,
      "loss": 0.3748,
      "step": 1168
    },
    {
      "epoch": 2.08,
      "learning_rate": 1.8729778310365489e-06,
      "loss": 0.3987,
      "step": 1169
    },
    {
      "epoch": 2.08,
      "learning_rate": 1.8693828639904134e-06,
      "loss": 0.3535,
      "step": 1170
    },
    {
      "epoch": 2.08,
      "learning_rate": 1.8657878969442782e-06,
      "loss": 0.3797,
      "step": 1171
    },
    {
      "epoch": 2.08,
      "learning_rate": 1.8621929298981425e-06,
      "loss": 0.3585,
      "step": 1172
    },
    {
      "epoch": 2.08,
      "learning_rate": 1.8585979628520073e-06,
      "loss": 0.4021,
      "step": 1173
    },
    {
      "epoch": 2.09,
      "learning_rate": 1.855002995805872e-06,
      "loss": 0.3767,
      "step": 1174
    },
    {
      "epoch": 2.09,
      "learning_rate": 1.8514080287597363e-06,
      "loss": 0.3591,
      "step": 1175
    },
    {
      "epoch": 2.09,
      "learning_rate": 1.847813061713601e-06,
      "loss": 0.3696,
      "step": 1176
    },
    {
      "epoch": 2.09,
      "learning_rate": 1.8442180946674654e-06,
      "loss": 0.3778,
      "step": 1177
    },
    {
      "epoch": 2.09,
      "learning_rate": 1.8406231276213302e-06,
      "loss": 0.3843,
      "step": 1178
    },
    {
      "epoch": 2.09,
      "learning_rate": 1.8370281605751947e-06,
      "loss": 0.4078,
      "step": 1179
    },
    {
      "epoch": 2.1,
      "learning_rate": 1.8334331935290594e-06,
      "loss": 0.3697,
      "step": 1180
    },
    {
      "epoch": 2.1,
      "learning_rate": 1.8298382264829242e-06,
      "loss": 0.3688,
      "step": 1181
    },
    {
      "epoch": 2.1,
      "learning_rate": 1.8262432594367885e-06,
      "loss": 0.3795,
      "step": 1182
    },
    {
      "epoch": 2.1,
      "learning_rate": 1.8226482923906533e-06,
      "loss": 0.3862,
      "step": 1183
    },
    {
      "epoch": 2.1,
      "learning_rate": 1.8190533253445176e-06,
      "loss": 0.3943,
      "step": 1184
    },
    {
      "epoch": 2.1,
      "learning_rate": 1.8154583582983823e-06,
      "loss": 0.3758,
      "step": 1185
    },
    {
      "epoch": 2.11,
      "learning_rate": 1.8118633912522467e-06,
      "loss": 0.3771,
      "step": 1186
    },
    {
      "epoch": 2.11,
      "learning_rate": 1.8082684242061114e-06,
      "loss": 0.3867,
      "step": 1187
    },
    {
      "epoch": 2.11,
      "learning_rate": 1.8046734571599762e-06,
      "loss": 0.3993,
      "step": 1188
    },
    {
      "epoch": 2.11,
      "learning_rate": 1.8010784901138407e-06,
      "loss": 0.3802,
      "step": 1189
    },
    {
      "epoch": 2.11,
      "learning_rate": 1.7974835230677055e-06,
      "loss": 0.4057,
      "step": 1190
    },
    {
      "epoch": 2.12,
      "learning_rate": 1.7938885560215698e-06,
      "loss": 0.369,
      "step": 1191
    },
    {
      "epoch": 2.12,
      "learning_rate": 1.7902935889754345e-06,
      "loss": 0.3552,
      "step": 1192
    },
    {
      "epoch": 2.12,
      "learning_rate": 1.7866986219292989e-06,
      "loss": 0.3775,
      "step": 1193
    },
    {
      "epoch": 2.12,
      "learning_rate": 1.7831036548831636e-06,
      "loss": 0.3756,
      "step": 1194
    },
    {
      "epoch": 2.12,
      "learning_rate": 1.7795086878370284e-06,
      "loss": 0.3768,
      "step": 1195
    },
    {
      "epoch": 2.12,
      "learning_rate": 1.7759137207908927e-06,
      "loss": 0.3957,
      "step": 1196
    },
    {
      "epoch": 2.13,
      "learning_rate": 1.7723187537447574e-06,
      "loss": 0.4085,
      "step": 1197
    },
    {
      "epoch": 2.13,
      "learning_rate": 1.768723786698622e-06,
      "loss": 0.348,
      "step": 1198
    },
    {
      "epoch": 2.13,
      "learning_rate": 1.7651288196524867e-06,
      "loss": 0.3953,
      "step": 1199
    },
    {
      "epoch": 2.13,
      "learning_rate": 1.761533852606351e-06,
      "loss": 0.378,
      "step": 1200
    },
    {
      "epoch": 2.13,
      "learning_rate": 1.7579388855602158e-06,
      "loss": 0.3601,
      "step": 1201
    },
    {
      "epoch": 2.13,
      "learning_rate": 1.7543439185140805e-06,
      "loss": 0.3766,
      "step": 1202
    },
    {
      "epoch": 2.14,
      "learning_rate": 1.7507489514679449e-06,
      "loss": 0.3893,
      "step": 1203
    },
    {
      "epoch": 2.14,
      "learning_rate": 1.7471539844218096e-06,
      "loss": 0.3732,
      "step": 1204
    },
    {
      "epoch": 2.14,
      "learning_rate": 1.743559017375674e-06,
      "loss": 0.3958,
      "step": 1205
    },
    {
      "epoch": 2.14,
      "learning_rate": 1.7399640503295387e-06,
      "loss": 0.3883,
      "step": 1206
    },
    {
      "epoch": 2.14,
      "learning_rate": 1.7363690832834032e-06,
      "loss": 0.4191,
      "step": 1207
    },
    {
      "epoch": 2.15,
      "learning_rate": 1.732774116237268e-06,
      "loss": 0.4085,
      "step": 1208
    },
    {
      "epoch": 2.15,
      "learning_rate": 1.7291791491911325e-06,
      "loss": 0.3633,
      "step": 1209
    },
    {
      "epoch": 2.15,
      "learning_rate": 1.725584182144997e-06,
      "loss": 0.4038,
      "step": 1210
    },
    {
      "epoch": 2.15,
      "learning_rate": 1.7219892150988618e-06,
      "loss": 0.3516,
      "step": 1211
    },
    {
      "epoch": 2.15,
      "learning_rate": 1.7183942480527261e-06,
      "loss": 0.3748,
      "step": 1212
    },
    {
      "epoch": 2.15,
      "learning_rate": 1.7147992810065909e-06,
      "loss": 0.386,
      "step": 1213
    },
    {
      "epoch": 2.16,
      "learning_rate": 1.7112043139604552e-06,
      "loss": 0.3697,
      "step": 1214
    },
    {
      "epoch": 2.16,
      "learning_rate": 1.70760934691432e-06,
      "loss": 0.3897,
      "step": 1215
    },
    {
      "epoch": 2.16,
      "learning_rate": 1.7040143798681845e-06,
      "loss": 0.3766,
      "step": 1216
    },
    {
      "epoch": 2.16,
      "learning_rate": 1.7004194128220492e-06,
      "loss": 0.3784,
      "step": 1217
    },
    {
      "epoch": 2.16,
      "learning_rate": 1.6968244457759138e-06,
      "loss": 0.3998,
      "step": 1218
    },
    {
      "epoch": 2.17,
      "learning_rate": 1.6932294787297783e-06,
      "loss": 0.3884,
      "step": 1219
    },
    {
      "epoch": 2.17,
      "learning_rate": 1.689634511683643e-06,
      "loss": 0.3845,
      "step": 1220
    },
    {
      "epoch": 2.17,
      "learning_rate": 1.6860395446375074e-06,
      "loss": 0.4006,
      "step": 1221
    },
    {
      "epoch": 2.17,
      "learning_rate": 1.6824445775913721e-06,
      "loss": 0.3698,
      "step": 1222
    },
    {
      "epoch": 2.17,
      "learning_rate": 1.6788496105452365e-06,
      "loss": 0.3667,
      "step": 1223
    },
    {
      "epoch": 2.17,
      "learning_rate": 1.6752546434991012e-06,
      "loss": 0.3762,
      "step": 1224
    },
    {
      "epoch": 2.18,
      "learning_rate": 1.671659676452966e-06,
      "loss": 0.3845,
      "step": 1225
    },
    {
      "epoch": 2.18,
      "learning_rate": 1.6680647094068305e-06,
      "loss": 0.3782,
      "step": 1226
    },
    {
      "epoch": 2.18,
      "learning_rate": 1.6644697423606953e-06,
      "loss": 0.381,
      "step": 1227
    },
    {
      "epoch": 2.18,
      "learning_rate": 1.6608747753145596e-06,
      "loss": 0.3852,
      "step": 1228
    },
    {
      "epoch": 2.18,
      "learning_rate": 1.6572798082684243e-06,
      "loss": 0.409,
      "step": 1229
    },
    {
      "epoch": 2.18,
      "learning_rate": 1.6536848412222887e-06,
      "loss": 0.4039,
      "step": 1230
    },
    {
      "epoch": 2.19,
      "learning_rate": 1.6500898741761534e-06,
      "loss": 0.3915,
      "step": 1231
    },
    {
      "epoch": 2.19,
      "learning_rate": 1.6464949071300182e-06,
      "loss": 0.3991,
      "step": 1232
    },
    {
      "epoch": 2.19,
      "learning_rate": 1.6428999400838825e-06,
      "loss": 0.386,
      "step": 1233
    },
    {
      "epoch": 2.19,
      "learning_rate": 1.6393049730377472e-06,
      "loss": 0.3792,
      "step": 1234
    },
    {
      "epoch": 2.19,
      "learning_rate": 1.6357100059916118e-06,
      "loss": 0.3914,
      "step": 1235
    },
    {
      "epoch": 2.2,
      "learning_rate": 1.6321150389454765e-06,
      "loss": 0.3955,
      "step": 1236
    },
    {
      "epoch": 2.2,
      "learning_rate": 1.6285200718993409e-06,
      "loss": 0.3958,
      "step": 1237
    },
    {
      "epoch": 2.2,
      "learning_rate": 1.6249251048532056e-06,
      "loss": 0.3759,
      "step": 1238
    },
    {
      "epoch": 2.2,
      "learning_rate": 1.6213301378070704e-06,
      "loss": 0.3946,
      "step": 1239
    },
    {
      "epoch": 2.2,
      "learning_rate": 1.6177351707609347e-06,
      "loss": 0.3975,
      "step": 1240
    },
    {
      "epoch": 2.2,
      "learning_rate": 1.6141402037147994e-06,
      "loss": 0.3819,
      "step": 1241
    },
    {
      "epoch": 2.21,
      "learning_rate": 1.6105452366686638e-06,
      "loss": 0.3853,
      "step": 1242
    },
    {
      "epoch": 2.21,
      "learning_rate": 1.6069502696225285e-06,
      "loss": 0.3692,
      "step": 1243
    },
    {
      "epoch": 2.21,
      "learning_rate": 1.603355302576393e-06,
      "loss": 0.3661,
      "step": 1244
    },
    {
      "epoch": 2.21,
      "learning_rate": 1.5997603355302578e-06,
      "loss": 0.378,
      "step": 1245
    },
    {
      "epoch": 2.21,
      "learning_rate": 1.5961653684841223e-06,
      "loss": 0.4056,
      "step": 1246
    },
    {
      "epoch": 2.21,
      "learning_rate": 1.5925704014379869e-06,
      "loss": 0.3921,
      "step": 1247
    },
    {
      "epoch": 2.22,
      "learning_rate": 1.5889754343918516e-06,
      "loss": 0.366,
      "step": 1248
    },
    {
      "epoch": 2.22,
      "learning_rate": 1.585380467345716e-06,
      "loss": 0.374,
      "step": 1249
    },
    {
      "epoch": 2.22,
      "learning_rate": 1.5817855002995807e-06,
      "loss": 0.3851,
      "step": 1250
    },
    {
      "epoch": 2.22,
      "learning_rate": 1.578190533253445e-06,
      "loss": 0.4058,
      "step": 1251
    },
    {
      "epoch": 2.22,
      "learning_rate": 1.5745955662073098e-06,
      "loss": 0.3931,
      "step": 1252
    },
    {
      "epoch": 2.23,
      "learning_rate": 1.5710005991611745e-06,
      "loss": 0.3914,
      "step": 1253
    },
    {
      "epoch": 2.23,
      "learning_rate": 1.567405632115039e-06,
      "loss": 0.3927,
      "step": 1254
    },
    {
      "epoch": 2.23,
      "learning_rate": 1.5638106650689036e-06,
      "loss": 0.3705,
      "step": 1255
    },
    {
      "epoch": 2.23,
      "learning_rate": 1.5602156980227681e-06,
      "loss": 0.4131,
      "step": 1256
    },
    {
      "epoch": 2.23,
      "learning_rate": 1.5566207309766329e-06,
      "loss": 0.3803,
      "step": 1257
    },
    {
      "epoch": 2.23,
      "learning_rate": 1.5530257639304972e-06,
      "loss": 0.3561,
      "step": 1258
    },
    {
      "epoch": 2.24,
      "learning_rate": 1.549430796884362e-06,
      "loss": 0.3786,
      "step": 1259
    },
    {
      "epoch": 2.24,
      "learning_rate": 1.5458358298382267e-06,
      "loss": 0.3689,
      "step": 1260
    },
    {
      "epoch": 2.24,
      "learning_rate": 1.542240862792091e-06,
      "loss": 0.3865,
      "step": 1261
    },
    {
      "epoch": 2.24,
      "learning_rate": 1.5386458957459558e-06,
      "loss": 0.3607,
      "step": 1262
    },
    {
      "epoch": 2.24,
      "learning_rate": 1.5350509286998203e-06,
      "loss": 0.375,
      "step": 1263
    },
    {
      "epoch": 2.25,
      "learning_rate": 1.5314559616536849e-06,
      "loss": 0.3895,
      "step": 1264
    },
    {
      "epoch": 2.25,
      "learning_rate": 1.5278609946075494e-06,
      "loss": 0.3832,
      "step": 1265
    },
    {
      "epoch": 2.25,
      "learning_rate": 1.5242660275614141e-06,
      "loss": 0.3954,
      "step": 1266
    },
    {
      "epoch": 2.25,
      "learning_rate": 1.5206710605152785e-06,
      "loss": 0.3985,
      "step": 1267
    },
    {
      "epoch": 2.25,
      "learning_rate": 1.5170760934691432e-06,
      "loss": 0.3944,
      "step": 1268
    },
    {
      "epoch": 2.25,
      "learning_rate": 1.513481126423008e-06,
      "loss": 0.365,
      "step": 1269
    },
    {
      "epoch": 2.26,
      "learning_rate": 1.5098861593768723e-06,
      "loss": 0.3923,
      "step": 1270
    },
    {
      "epoch": 2.26,
      "learning_rate": 1.506291192330737e-06,
      "loss": 0.4005,
      "step": 1271
    },
    {
      "epoch": 2.26,
      "learning_rate": 1.5026962252846016e-06,
      "loss": 0.3782,
      "step": 1272
    },
    {
      "epoch": 2.26,
      "learning_rate": 1.4991012582384661e-06,
      "loss": 0.3849,
      "step": 1273
    },
    {
      "epoch": 2.26,
      "learning_rate": 1.4955062911923309e-06,
      "loss": 0.3749,
      "step": 1274
    },
    {
      "epoch": 2.26,
      "learning_rate": 1.4919113241461954e-06,
      "loss": 0.3955,
      "step": 1275
    },
    {
      "epoch": 2.27,
      "learning_rate": 1.48831635710006e-06,
      "loss": 0.3781,
      "step": 1276
    },
    {
      "epoch": 2.27,
      "learning_rate": 1.4847213900539245e-06,
      "loss": 0.383,
      "step": 1277
    },
    {
      "epoch": 2.27,
      "learning_rate": 1.481126423007789e-06,
      "loss": 0.379,
      "step": 1278
    },
    {
      "epoch": 2.27,
      "learning_rate": 1.4775314559616538e-06,
      "loss": 0.3646,
      "step": 1279
    },
    {
      "epoch": 2.27,
      "learning_rate": 1.4739364889155183e-06,
      "loss": 0.4073,
      "step": 1280
    },
    {
      "epoch": 2.28,
      "learning_rate": 1.4703415218693828e-06,
      "loss": 0.3815,
      "step": 1281
    },
    {
      "epoch": 2.28,
      "learning_rate": 1.4667465548232474e-06,
      "loss": 0.3898,
      "step": 1282
    },
    {
      "epoch": 2.28,
      "learning_rate": 1.4631515877771121e-06,
      "loss": 0.383,
      "step": 1283
    },
    {
      "epoch": 2.28,
      "learning_rate": 1.4595566207309767e-06,
      "loss": 0.3815,
      "step": 1284
    },
    {
      "epoch": 2.28,
      "learning_rate": 1.4559616536848412e-06,
      "loss": 0.3785,
      "step": 1285
    },
    {
      "epoch": 2.28,
      "learning_rate": 1.452366686638706e-06,
      "loss": 0.3936,
      "step": 1286
    },
    {
      "epoch": 2.29,
      "learning_rate": 1.4487717195925705e-06,
      "loss": 0.389,
      "step": 1287
    },
    {
      "epoch": 2.29,
      "learning_rate": 1.445176752546435e-06,
      "loss": 0.3477,
      "step": 1288
    },
    {
      "epoch": 2.29,
      "learning_rate": 1.4415817855002996e-06,
      "loss": 0.3881,
      "step": 1289
    },
    {
      "epoch": 2.29,
      "learning_rate": 1.4379868184541641e-06,
      "loss": 0.3872,
      "step": 1290
    },
    {
      "epoch": 2.29,
      "learning_rate": 1.4343918514080289e-06,
      "loss": 0.3953,
      "step": 1291
    },
    {
      "epoch": 2.29,
      "learning_rate": 1.4307968843618934e-06,
      "loss": 0.3643,
      "step": 1292
    },
    {
      "epoch": 2.3,
      "learning_rate": 1.4272019173157581e-06,
      "loss": 0.398,
      "step": 1293
    },
    {
      "epoch": 2.3,
      "learning_rate": 1.4236069502696227e-06,
      "loss": 0.3761,
      "step": 1294
    },
    {
      "epoch": 2.3,
      "learning_rate": 1.4200119832234872e-06,
      "loss": 0.3822,
      "step": 1295
    },
    {
      "epoch": 2.3,
      "learning_rate": 1.4164170161773518e-06,
      "loss": 0.3847,
      "step": 1296
    },
    {
      "epoch": 2.3,
      "learning_rate": 1.4128220491312163e-06,
      "loss": 0.3913,
      "step": 1297
    },
    {
      "epoch": 2.31,
      "learning_rate": 1.4092270820850808e-06,
      "loss": 0.3911,
      "step": 1298
    },
    {
      "epoch": 2.31,
      "learning_rate": 1.4056321150389454e-06,
      "loss": 0.4136,
      "step": 1299
    },
    {
      "epoch": 2.31,
      "learning_rate": 1.4020371479928101e-06,
      "loss": 0.3746,
      "step": 1300
    },
    {
      "epoch": 2.31,
      "learning_rate": 1.3984421809466747e-06,
      "loss": 0.3588,
      "step": 1301
    },
    {
      "epoch": 2.31,
      "learning_rate": 1.3948472139005394e-06,
      "loss": 0.3817,
      "step": 1302
    },
    {
      "epoch": 2.31,
      "learning_rate": 1.391252246854404e-06,
      "loss": 0.3898,
      "step": 1303
    },
    {
      "epoch": 2.32,
      "learning_rate": 1.3876572798082685e-06,
      "loss": 0.4131,
      "step": 1304
    },
    {
      "epoch": 2.32,
      "learning_rate": 1.384062312762133e-06,
      "loss": 0.3932,
      "step": 1305
    },
    {
      "epoch": 2.32,
      "learning_rate": 1.3804673457159976e-06,
      "loss": 0.3614,
      "step": 1306
    },
    {
      "epoch": 2.32,
      "learning_rate": 1.376872378669862e-06,
      "loss": 0.375,
      "step": 1307
    },
    {
      "epoch": 2.32,
      "learning_rate": 1.3732774116237268e-06,
      "loss": 0.3691,
      "step": 1308
    },
    {
      "epoch": 2.33,
      "learning_rate": 1.3696824445775914e-06,
      "loss": 0.379,
      "step": 1309
    },
    {
      "epoch": 2.33,
      "learning_rate": 1.366087477531456e-06,
      "loss": 0.3837,
      "step": 1310
    },
    {
      "epoch": 2.33,
      "learning_rate": 1.3624925104853207e-06,
      "loss": 0.3866,
      "step": 1311
    },
    {
      "epoch": 2.33,
      "learning_rate": 1.3588975434391852e-06,
      "loss": 0.3718,
      "step": 1312
    },
    {
      "epoch": 2.33,
      "learning_rate": 1.3553025763930498e-06,
      "loss": 0.3915,
      "step": 1313
    },
    {
      "epoch": 2.33,
      "learning_rate": 1.3517076093469143e-06,
      "loss": 0.4006,
      "step": 1314
    },
    {
      "epoch": 2.34,
      "learning_rate": 1.348112642300779e-06,
      "loss": 0.3784,
      "step": 1315
    },
    {
      "epoch": 2.34,
      "learning_rate": 1.3445176752546436e-06,
      "loss": 0.3819,
      "step": 1316
    },
    {
      "epoch": 2.34,
      "learning_rate": 1.3409227082085081e-06,
      "loss": 0.3975,
      "step": 1317
    },
    {
      "epoch": 2.34,
      "learning_rate": 1.3373277411623727e-06,
      "loss": 0.4042,
      "step": 1318
    },
    {
      "epoch": 2.34,
      "learning_rate": 1.3337327741162372e-06,
      "loss": 0.3867,
      "step": 1319
    },
    {
      "epoch": 2.34,
      "learning_rate": 1.330137807070102e-06,
      "loss": 0.389,
      "step": 1320
    },
    {
      "epoch": 2.35,
      "learning_rate": 1.3265428400239665e-06,
      "loss": 0.384,
      "step": 1321
    },
    {
      "epoch": 2.35,
      "learning_rate": 1.3229478729778312e-06,
      "loss": 0.3988,
      "step": 1322
    },
    {
      "epoch": 2.35,
      "learning_rate": 1.3193529059316958e-06,
      "loss": 0.3656,
      "step": 1323
    },
    {
      "epoch": 2.35,
      "learning_rate": 1.3157579388855603e-06,
      "loss": 0.3843,
      "step": 1324
    },
    {
      "epoch": 2.35,
      "learning_rate": 1.3121629718394248e-06,
      "loss": 0.3816,
      "step": 1325
    },
    {
      "epoch": 2.36,
      "learning_rate": 1.3085680047932894e-06,
      "loss": 0.3821,
      "step": 1326
    },
    {
      "epoch": 2.36,
      "learning_rate": 1.304973037747154e-06,
      "loss": 0.3755,
      "step": 1327
    },
    {
      "epoch": 2.36,
      "learning_rate": 1.3013780707010185e-06,
      "loss": 0.3694,
      "step": 1328
    },
    {
      "epoch": 2.36,
      "learning_rate": 1.2977831036548832e-06,
      "loss": 0.3647,
      "step": 1329
    },
    {
      "epoch": 2.36,
      "learning_rate": 1.294188136608748e-06,
      "loss": 0.3685,
      "step": 1330
    },
    {
      "epoch": 2.36,
      "learning_rate": 1.2905931695626125e-06,
      "loss": 0.3939,
      "step": 1331
    },
    {
      "epoch": 2.37,
      "learning_rate": 1.286998202516477e-06,
      "loss": 0.3821,
      "step": 1332
    },
    {
      "epoch": 2.37,
      "learning_rate": 1.2834032354703416e-06,
      "loss": 0.4043,
      "step": 1333
    },
    {
      "epoch": 2.37,
      "learning_rate": 1.279808268424206e-06,
      "loss": 0.3787,
      "step": 1334
    },
    {
      "epoch": 2.37,
      "learning_rate": 1.2762133013780706e-06,
      "loss": 0.3687,
      "step": 1335
    },
    {
      "epoch": 2.37,
      "learning_rate": 1.2726183343319352e-06,
      "loss": 0.3749,
      "step": 1336
    },
    {
      "epoch": 2.37,
      "learning_rate": 1.2690233672858e-06,
      "loss": 0.3897,
      "step": 1337
    },
    {
      "epoch": 2.38,
      "learning_rate": 1.2654284002396645e-06,
      "loss": 0.3888,
      "step": 1338
    },
    {
      "epoch": 2.38,
      "learning_rate": 1.2618334331935292e-06,
      "loss": 0.3519,
      "step": 1339
    },
    {
      "epoch": 2.38,
      "learning_rate": 1.2582384661473938e-06,
      "loss": 0.3786,
      "step": 1340
    },
    {
      "epoch": 2.38,
      "learning_rate": 1.2546434991012583e-06,
      "loss": 0.4022,
      "step": 1341
    },
    {
      "epoch": 2.38,
      "learning_rate": 1.2510485320551228e-06,
      "loss": 0.3579,
      "step": 1342
    },
    {
      "epoch": 2.39,
      "learning_rate": 1.2474535650089874e-06,
      "loss": 0.3645,
      "step": 1343
    },
    {
      "epoch": 2.39,
      "learning_rate": 1.2438585979628521e-06,
      "loss": 0.371,
      "step": 1344
    },
    {
      "epoch": 2.39,
      "learning_rate": 1.2402636309167167e-06,
      "loss": 0.3872,
      "step": 1345
    },
    {
      "epoch": 2.39,
      "learning_rate": 1.2366686638705812e-06,
      "loss": 0.4014,
      "step": 1346
    },
    {
      "epoch": 2.39,
      "learning_rate": 1.2330736968244457e-06,
      "loss": 0.367,
      "step": 1347
    },
    {
      "epoch": 2.39,
      "learning_rate": 1.2294787297783105e-06,
      "loss": 0.3742,
      "step": 1348
    },
    {
      "epoch": 2.4,
      "learning_rate": 1.225883762732175e-06,
      "loss": 0.3634,
      "step": 1349
    },
    {
      "epoch": 2.4,
      "learning_rate": 1.2222887956860396e-06,
      "loss": 0.3724,
      "step": 1350
    },
    {
      "epoch": 2.4,
      "learning_rate": 1.218693828639904e-06,
      "loss": 0.367,
      "step": 1351
    },
    {
      "epoch": 2.4,
      "learning_rate": 1.2150988615937688e-06,
      "loss": 0.3546,
      "step": 1352
    },
    {
      "epoch": 2.4,
      "learning_rate": 1.2115038945476334e-06,
      "loss": 0.3696,
      "step": 1353
    },
    {
      "epoch": 2.4,
      "learning_rate": 1.207908927501498e-06,
      "loss": 0.3987,
      "step": 1354
    },
    {
      "epoch": 2.41,
      "learning_rate": 1.2043139604553625e-06,
      "loss": 0.3611,
      "step": 1355
    },
    {
      "epoch": 2.41,
      "learning_rate": 1.200718993409227e-06,
      "loss": 0.3766,
      "step": 1356
    },
    {
      "epoch": 2.41,
      "learning_rate": 1.1971240263630917e-06,
      "loss": 0.3805,
      "step": 1357
    },
    {
      "epoch": 2.41,
      "learning_rate": 1.1935290593169563e-06,
      "loss": 0.3751,
      "step": 1358
    },
    {
      "epoch": 2.41,
      "learning_rate": 1.189934092270821e-06,
      "loss": 0.3884,
      "step": 1359
    },
    {
      "epoch": 2.42,
      "learning_rate": 1.1863391252246856e-06,
      "loss": 0.3831,
      "step": 1360
    },
    {
      "epoch": 2.42,
      "learning_rate": 1.1827441581785501e-06,
      "loss": 0.3678,
      "step": 1361
    },
    {
      "epoch": 2.42,
      "learning_rate": 1.1791491911324146e-06,
      "loss": 0.4019,
      "step": 1362
    },
    {
      "epoch": 2.42,
      "learning_rate": 1.1755542240862792e-06,
      "loss": 0.4072,
      "step": 1363
    },
    {
      "epoch": 2.42,
      "learning_rate": 1.1719592570401437e-06,
      "loss": 0.3665,
      "step": 1364
    },
    {
      "epoch": 2.42,
      "learning_rate": 1.1683642899940083e-06,
      "loss": 0.3896,
      "step": 1365
    },
    {
      "epoch": 2.43,
      "learning_rate": 1.164769322947873e-06,
      "loss": 0.3716,
      "step": 1366
    },
    {
      "epoch": 2.43,
      "learning_rate": 1.1611743559017375e-06,
      "loss": 0.4011,
      "step": 1367
    },
    {
      "epoch": 2.43,
      "learning_rate": 1.1575793888556023e-06,
      "loss": 0.3941,
      "step": 1368
    },
    {
      "epoch": 2.43,
      "learning_rate": 1.1539844218094668e-06,
      "loss": 0.3912,
      "step": 1369
    },
    {
      "epoch": 2.43,
      "learning_rate": 1.1503894547633314e-06,
      "loss": 0.3958,
      "step": 1370
    },
    {
      "epoch": 2.44,
      "learning_rate": 1.146794487717196e-06,
      "loss": 0.3927,
      "step": 1371
    },
    {
      "epoch": 2.44,
      "learning_rate": 1.1431995206710604e-06,
      "loss": 0.3925,
      "step": 1372
    },
    {
      "epoch": 2.44,
      "learning_rate": 1.1396045536249252e-06,
      "loss": 0.3736,
      "step": 1373
    },
    {
      "epoch": 2.44,
      "learning_rate": 1.1360095865787897e-06,
      "loss": 0.3956,
      "step": 1374
    },
    {
      "epoch": 2.44,
      "learning_rate": 1.1324146195326543e-06,
      "loss": 0.3702,
      "step": 1375
    },
    {
      "epoch": 2.44,
      "learning_rate": 1.1288196524865188e-06,
      "loss": 0.3729,
      "step": 1376
    },
    {
      "epoch": 2.45,
      "learning_rate": 1.1252246854403836e-06,
      "loss": 0.3713,
      "step": 1377
    },
    {
      "epoch": 2.45,
      "learning_rate": 1.121629718394248e-06,
      "loss": 0.4047,
      "step": 1378
    },
    {
      "epoch": 2.45,
      "learning_rate": 1.1180347513481126e-06,
      "loss": 0.378,
      "step": 1379
    },
    {
      "epoch": 2.45,
      "learning_rate": 1.1144397843019772e-06,
      "loss": 0.4085,
      "step": 1380
    },
    {
      "epoch": 2.45,
      "learning_rate": 1.110844817255842e-06,
      "loss": 0.3505,
      "step": 1381
    },
    {
      "epoch": 2.45,
      "learning_rate": 1.1072498502097065e-06,
      "loss": 0.3884,
      "step": 1382
    },
    {
      "epoch": 2.46,
      "learning_rate": 1.103654883163571e-06,
      "loss": 0.373,
      "step": 1383
    },
    {
      "epoch": 2.46,
      "learning_rate": 1.1000599161174355e-06,
      "loss": 0.3997,
      "step": 1384
    },
    {
      "epoch": 2.46,
      "learning_rate": 1.0964649490713003e-06,
      "loss": 0.3655,
      "step": 1385
    },
    {
      "epoch": 2.46,
      "learning_rate": 1.0928699820251648e-06,
      "loss": 0.3943,
      "step": 1386
    },
    {
      "epoch": 2.46,
      "learning_rate": 1.0892750149790294e-06,
      "loss": 0.3814,
      "step": 1387
    },
    {
      "epoch": 2.47,
      "learning_rate": 1.0856800479328941e-06,
      "loss": 0.3955,
      "step": 1388
    },
    {
      "epoch": 2.47,
      "learning_rate": 1.0820850808867587e-06,
      "loss": 0.3905,
      "step": 1389
    },
    {
      "epoch": 2.47,
      "learning_rate": 1.0784901138406232e-06,
      "loss": 0.3803,
      "step": 1390
    },
    {
      "epoch": 2.47,
      "learning_rate": 1.0748951467944877e-06,
      "loss": 0.3784,
      "step": 1391
    },
    {
      "epoch": 2.47,
      "learning_rate": 1.0713001797483523e-06,
      "loss": 0.3857,
      "step": 1392
    },
    {
      "epoch": 2.47,
      "learning_rate": 1.0677052127022168e-06,
      "loss": 0.3759,
      "step": 1393
    },
    {
      "epoch": 2.48,
      "learning_rate": 1.0641102456560816e-06,
      "loss": 0.389,
      "step": 1394
    },
    {
      "epoch": 2.48,
      "learning_rate": 1.060515278609946e-06,
      "loss": 0.406,
      "step": 1395
    },
    {
      "epoch": 2.48,
      "learning_rate": 1.0569203115638108e-06,
      "loss": 0.4044,
      "step": 1396
    },
    {
      "epoch": 2.48,
      "learning_rate": 1.0533253445176754e-06,
      "loss": 0.3742,
      "step": 1397
    },
    {
      "epoch": 2.48,
      "learning_rate": 1.04973037747154e-06,
      "loss": 0.3801,
      "step": 1398
    },
    {
      "epoch": 2.48,
      "learning_rate": 1.0461354104254045e-06,
      "loss": 0.362,
      "step": 1399
    },
    {
      "epoch": 2.49,
      "learning_rate": 1.042540443379269e-06,
      "loss": 0.383,
      "step": 1400
    },
    {
      "epoch": 2.49,
      "learning_rate": 1.0389454763331335e-06,
      "loss": 0.3742,
      "step": 1401
    },
    {
      "epoch": 2.49,
      "learning_rate": 1.035350509286998e-06,
      "loss": 0.4001,
      "step": 1402
    },
    {
      "epoch": 2.49,
      "learning_rate": 1.0317555422408628e-06,
      "loss": 0.3656,
      "step": 1403
    },
    {
      "epoch": 2.49,
      "learning_rate": 1.0281605751947274e-06,
      "loss": 0.3913,
      "step": 1404
    },
    {
      "epoch": 2.5,
      "learning_rate": 1.024565608148592e-06,
      "loss": 0.3671,
      "step": 1405
    },
    {
      "epoch": 2.5,
      "learning_rate": 1.0209706411024566e-06,
      "loss": 0.3938,
      "step": 1406
    },
    {
      "epoch": 2.5,
      "learning_rate": 1.0173756740563212e-06,
      "loss": 0.3503,
      "step": 1407
    },
    {
      "epoch": 2.5,
      "learning_rate": 1.0137807070101857e-06,
      "loss": 0.4179,
      "step": 1408
    },
    {
      "epoch": 2.5,
      "learning_rate": 1.0101857399640503e-06,
      "loss": 0.3798,
      "step": 1409
    },
    {
      "epoch": 2.5,
      "learning_rate": 1.006590772917915e-06,
      "loss": 0.3855,
      "step": 1410
    },
    {
      "epoch": 2.51,
      "learning_rate": 1.0029958058717795e-06,
      "loss": 0.3968,
      "step": 1411
    },
    {
      "epoch": 2.51,
      "learning_rate": 9.99400838825644e-07,
      "loss": 0.3928,
      "step": 1412
    },
    {
      "epoch": 2.51,
      "learning_rate": 9.958058717795086e-07,
      "loss": 0.3869,
      "step": 1413
    },
    {
      "epoch": 2.51,
      "learning_rate": 9.922109047333734e-07,
      "loss": 0.3977,
      "step": 1414
    },
    {
      "epoch": 2.51,
      "learning_rate": 9.88615937687238e-07,
      "loss": 0.3768,
      "step": 1415
    },
    {
      "epoch": 2.52,
      "learning_rate": 9.850209706411024e-07,
      "loss": 0.3931,
      "step": 1416
    },
    {
      "epoch": 2.52,
      "learning_rate": 9.814260035949672e-07,
      "loss": 0.3752,
      "step": 1417
    },
    {
      "epoch": 2.52,
      "learning_rate": 9.778310365488317e-07,
      "loss": 0.3668,
      "step": 1418
    },
    {
      "epoch": 2.52,
      "learning_rate": 9.742360695026963e-07,
      "loss": 0.3889,
      "step": 1419
    },
    {
      "epoch": 2.52,
      "learning_rate": 9.706411024565608e-07,
      "loss": 0.3622,
      "step": 1420
    },
    {
      "epoch": 2.52,
      "learning_rate": 9.670461354104253e-07,
      "loss": 0.3974,
      "step": 1421
    },
    {
      "epoch": 2.53,
      "learning_rate": 9.634511683642899e-07,
      "loss": 0.3881,
      "step": 1422
    },
    {
      "epoch": 2.53,
      "learning_rate": 9.598562013181546e-07,
      "loss": 0.3845,
      "step": 1423
    },
    {
      "epoch": 2.53,
      "learning_rate": 9.562612342720194e-07,
      "loss": 0.4073,
      "step": 1424
    },
    {
      "epoch": 2.53,
      "learning_rate": 9.526662672258838e-07,
      "loss": 0.3742,
      "step": 1425
    },
    {
      "epoch": 2.53,
      "learning_rate": 9.490713001797485e-07,
      "loss": 0.3811,
      "step": 1426
    },
    {
      "epoch": 2.53,
      "learning_rate": 9.45476333133613e-07,
      "loss": 0.3758,
      "step": 1427
    },
    {
      "epoch": 2.54,
      "learning_rate": 9.418813660874775e-07,
      "loss": 0.3981,
      "step": 1428
    },
    {
      "epoch": 2.54,
      "learning_rate": 9.382863990413421e-07,
      "loss": 0.3845,
      "step": 1429
    },
    {
      "epoch": 2.54,
      "learning_rate": 9.346914319952067e-07,
      "loss": 0.3893,
      "step": 1430
    },
    {
      "epoch": 2.54,
      "learning_rate": 9.310964649490713e-07,
      "loss": 0.355,
      "step": 1431
    },
    {
      "epoch": 2.54,
      "learning_rate": 9.27501497902936e-07,
      "loss": 0.3973,
      "step": 1432
    },
    {
      "epoch": 2.55,
      "learning_rate": 9.239065308568005e-07,
      "loss": 0.3876,
      "step": 1433
    },
    {
      "epoch": 2.55,
      "learning_rate": 9.203115638106651e-07,
      "loss": 0.3628,
      "step": 1434
    },
    {
      "epoch": 2.55,
      "learning_rate": 9.167165967645297e-07,
      "loss": 0.3559,
      "step": 1435
    },
    {
      "epoch": 2.55,
      "learning_rate": 9.131216297183943e-07,
      "loss": 0.3947,
      "step": 1436
    },
    {
      "epoch": 2.55,
      "learning_rate": 9.095266626722588e-07,
      "loss": 0.371,
      "step": 1437
    },
    {
      "epoch": 2.55,
      "learning_rate": 9.059316956261233e-07,
      "loss": 0.3721,
      "step": 1438
    },
    {
      "epoch": 2.56,
      "learning_rate": 9.023367285799881e-07,
      "loss": 0.3789,
      "step": 1439
    },
    {
      "epoch": 2.56,
      "learning_rate": 8.987417615338527e-07,
      "loss": 0.3991,
      "step": 1440
    },
    {
      "epoch": 2.56,
      "learning_rate": 8.951467944877173e-07,
      "loss": 0.3944,
      "step": 1441
    },
    {
      "epoch": 2.56,
      "learning_rate": 8.915518274415818e-07,
      "loss": 0.3903,
      "step": 1442
    },
    {
      "epoch": 2.56,
      "learning_rate": 8.879568603954463e-07,
      "loss": 0.3844,
      "step": 1443
    },
    {
      "epoch": 2.56,
      "learning_rate": 8.84361893349311e-07,
      "loss": 0.3725,
      "step": 1444
    },
    {
      "epoch": 2.57,
      "learning_rate": 8.807669263031755e-07,
      "loss": 0.396,
      "step": 1445
    },
    {
      "epoch": 2.57,
      "learning_rate": 8.771719592570403e-07,
      "loss": 0.3962,
      "step": 1446
    },
    {
      "epoch": 2.57,
      "learning_rate": 8.735769922109048e-07,
      "loss": 0.3999,
      "step": 1447
    },
    {
      "epoch": 2.57,
      "learning_rate": 8.699820251647693e-07,
      "loss": 0.3708,
      "step": 1448
    },
    {
      "epoch": 2.57,
      "learning_rate": 8.66387058118634e-07,
      "loss": 0.373,
      "step": 1449
    },
    {
      "epoch": 2.58,
      "learning_rate": 8.627920910724985e-07,
      "loss": 0.388,
      "step": 1450
    },
    {
      "epoch": 2.58,
      "learning_rate": 8.591971240263631e-07,
      "loss": 0.3472,
      "step": 1451
    },
    {
      "epoch": 2.58,
      "learning_rate": 8.556021569802276e-07,
      "loss": 0.3789,
      "step": 1452
    },
    {
      "epoch": 2.58,
      "learning_rate": 8.520071899340922e-07,
      "loss": 0.3786,
      "step": 1453
    },
    {
      "epoch": 2.58,
      "learning_rate": 8.484122228879569e-07,
      "loss": 0.3902,
      "step": 1454
    },
    {
      "epoch": 2.58,
      "learning_rate": 8.448172558418215e-07,
      "loss": 0.4055,
      "step": 1455
    },
    {
      "epoch": 2.59,
      "learning_rate": 8.412222887956861e-07,
      "loss": 0.368,
      "step": 1456
    },
    {
      "epoch": 2.59,
      "learning_rate": 8.376273217495506e-07,
      "loss": 0.4093,
      "step": 1457
    },
    {
      "epoch": 2.59,
      "learning_rate": 8.340323547034153e-07,
      "loss": 0.3868,
      "step": 1458
    },
    {
      "epoch": 2.59,
      "learning_rate": 8.304373876572798e-07,
      "loss": 0.3895,
      "step": 1459
    },
    {
      "epoch": 2.59,
      "learning_rate": 8.268424206111443e-07,
      "loss": 0.3981,
      "step": 1460
    },
    {
      "epoch": 2.6,
      "learning_rate": 8.232474535650091e-07,
      "loss": 0.3935,
      "step": 1461
    },
    {
      "epoch": 2.6,
      "learning_rate": 8.196524865188736e-07,
      "loss": 0.3512,
      "step": 1462
    },
    {
      "epoch": 2.6,
      "learning_rate": 8.160575194727383e-07,
      "loss": 0.4029,
      "step": 1463
    },
    {
      "epoch": 2.6,
      "learning_rate": 8.124625524266028e-07,
      "loss": 0.358,
      "step": 1464
    },
    {
      "epoch": 2.6,
      "learning_rate": 8.088675853804673e-07,
      "loss": 0.3713,
      "step": 1465
    },
    {
      "epoch": 2.6,
      "learning_rate": 8.052726183343319e-07,
      "loss": 0.3837,
      "step": 1466
    },
    {
      "epoch": 2.61,
      "learning_rate": 8.016776512881965e-07,
      "loss": 0.3939,
      "step": 1467
    },
    {
      "epoch": 2.61,
      "learning_rate": 7.980826842420612e-07,
      "loss": 0.3887,
      "step": 1468
    },
    {
      "epoch": 2.61,
      "learning_rate": 7.944877171959258e-07,
      "loss": 0.3964,
      "step": 1469
    },
    {
      "epoch": 2.61,
      "learning_rate": 7.908927501497903e-07,
      "loss": 0.4059,
      "step": 1470
    },
    {
      "epoch": 2.61,
      "learning_rate": 7.872977831036549e-07,
      "loss": 0.4006,
      "step": 1471
    },
    {
      "epoch": 2.61,
      "learning_rate": 7.837028160575195e-07,
      "loss": 0.3745,
      "step": 1472
    },
    {
      "epoch": 2.62,
      "learning_rate": 7.801078490113841e-07,
      "loss": 0.3708,
      "step": 1473
    },
    {
      "epoch": 2.62,
      "learning_rate": 7.765128819652486e-07,
      "loss": 0.387,
      "step": 1474
    },
    {
      "epoch": 2.62,
      "learning_rate": 7.729179149191134e-07,
      "loss": 0.3714,
      "step": 1475
    },
    {
      "epoch": 2.62,
      "learning_rate": 7.693229478729779e-07,
      "loss": 0.3772,
      "step": 1476
    },
    {
      "epoch": 2.62,
      "learning_rate": 7.657279808268424e-07,
      "loss": 0.398,
      "step": 1477
    },
    {
      "epoch": 2.63,
      "learning_rate": 7.621330137807071e-07,
      "loss": 0.3733,
      "step": 1478
    },
    {
      "epoch": 2.63,
      "learning_rate": 7.585380467345716e-07,
      "loss": 0.381,
      "step": 1479
    },
    {
      "epoch": 2.63,
      "learning_rate": 7.549430796884361e-07,
      "loss": 0.3771,
      "step": 1480
    },
    {
      "epoch": 2.63,
      "learning_rate": 7.513481126423008e-07,
      "loss": 0.4006,
      "step": 1481
    },
    {
      "epoch": 2.63,
      "learning_rate": 7.477531455961654e-07,
      "loss": 0.3774,
      "step": 1482
    },
    {
      "epoch": 2.63,
      "learning_rate": 7.4415817855003e-07,
      "loss": 0.3845,
      "step": 1483
    },
    {
      "epoch": 2.64,
      "learning_rate": 7.405632115038945e-07,
      "loss": 0.3757,
      "step": 1484
    },
    {
      "epoch": 2.64,
      "learning_rate": 7.369682444577592e-07,
      "loss": 0.3898,
      "step": 1485
    },
    {
      "epoch": 2.64,
      "learning_rate": 7.333732774116237e-07,
      "loss": 0.3608,
      "step": 1486
    },
    {
      "epoch": 2.64,
      "learning_rate": 7.297783103654883e-07,
      "loss": 0.3815,
      "step": 1487
    },
    {
      "epoch": 2.64,
      "learning_rate": 7.26183343319353e-07,
      "loss": 0.3798,
      "step": 1488
    },
    {
      "epoch": 2.64,
      "learning_rate": 7.225883762732175e-07,
      "loss": 0.3696,
      "step": 1489
    },
    {
      "epoch": 2.65,
      "learning_rate": 7.189934092270821e-07,
      "loss": 0.3855,
      "step": 1490
    },
    {
      "epoch": 2.65,
      "learning_rate": 7.153984421809467e-07,
      "loss": 0.3798,
      "step": 1491
    },
    {
      "epoch": 2.65,
      "learning_rate": 7.118034751348113e-07,
      "loss": 0.3957,
      "step": 1492
    },
    {
      "epoch": 2.65,
      "learning_rate": 7.082085080886759e-07,
      "loss": 0.4118,
      "step": 1493
    },
    {
      "epoch": 2.65,
      "learning_rate": 7.046135410425404e-07,
      "loss": 0.4045,
      "step": 1494
    },
    {
      "epoch": 2.66,
      "learning_rate": 7.010185739964051e-07,
      "loss": 0.3813,
      "step": 1495
    },
    {
      "epoch": 2.66,
      "learning_rate": 6.974236069502697e-07,
      "loss": 0.3962,
      "step": 1496
    },
    {
      "epoch": 2.66,
      "learning_rate": 6.938286399041342e-07,
      "loss": 0.3944,
      "step": 1497
    },
    {
      "epoch": 2.66,
      "learning_rate": 6.902336728579988e-07,
      "loss": 0.412,
      "step": 1498
    },
    {
      "epoch": 2.66,
      "learning_rate": 6.866387058118634e-07,
      "loss": 0.3843,
      "step": 1499
    },
    {
      "epoch": 2.66,
      "learning_rate": 6.83043738765728e-07,
      "loss": 0.3546,
      "step": 1500
    },
    {
      "epoch": 2.67,
      "learning_rate": 6.794487717195926e-07,
      "loss": 0.4036,
      "step": 1501
    },
    {
      "epoch": 2.67,
      "learning_rate": 6.758538046734571e-07,
      "loss": 0.3988,
      "step": 1502
    },
    {
      "epoch": 2.67,
      "learning_rate": 6.722588376273218e-07,
      "loss": 0.3828,
      "step": 1503
    },
    {
      "epoch": 2.67,
      "learning_rate": 6.686638705811863e-07,
      "loss": 0.3736,
      "step": 1504
    },
    {
      "epoch": 2.67,
      "learning_rate": 6.65068903535051e-07,
      "loss": 0.4018,
      "step": 1505
    },
    {
      "epoch": 2.67,
      "learning_rate": 6.614739364889156e-07,
      "loss": 0.373,
      "step": 1506
    },
    {
      "epoch": 2.68,
      "learning_rate": 6.578789694427802e-07,
      "loss": 0.3866,
      "step": 1507
    },
    {
      "epoch": 2.68,
      "learning_rate": 6.542840023966447e-07,
      "loss": 0.3727,
      "step": 1508
    },
    {
      "epoch": 2.68,
      "learning_rate": 6.506890353505092e-07,
      "loss": 0.3769,
      "step": 1509
    },
    {
      "epoch": 2.68,
      "learning_rate": 6.47094068304374e-07,
      "loss": 0.3846,
      "step": 1510
    },
    {
      "epoch": 2.68,
      "learning_rate": 6.434991012582385e-07,
      "loss": 0.3853,
      "step": 1511
    },
    {
      "epoch": 2.69,
      "learning_rate": 6.39904134212103e-07,
      "loss": 0.3675,
      "step": 1512
    },
    {
      "epoch": 2.69,
      "learning_rate": 6.363091671659676e-07,
      "loss": 0.3713,
      "step": 1513
    },
    {
      "epoch": 2.69,
      "learning_rate": 6.327142001198322e-07,
      "loss": 0.3716,
      "step": 1514
    },
    {
      "epoch": 2.69,
      "learning_rate": 6.291192330736969e-07,
      "loss": 0.3979,
      "step": 1515
    },
    {
      "epoch": 2.69,
      "learning_rate": 6.255242660275614e-07,
      "loss": 0.4059,
      "step": 1516
    },
    {
      "epoch": 2.69,
      "learning_rate": 6.219292989814261e-07,
      "loss": 0.3897,
      "step": 1517
    },
    {
      "epoch": 2.7,
      "learning_rate": 6.183343319352906e-07,
      "loss": 0.3806,
      "step": 1518
    },
    {
      "epoch": 2.7,
      "learning_rate": 6.147393648891552e-07,
      "loss": 0.3926,
      "step": 1519
    },
    {
      "epoch": 2.7,
      "learning_rate": 6.111443978430198e-07,
      "loss": 0.3624,
      "step": 1520
    },
    {
      "epoch": 2.7,
      "learning_rate": 6.075494307968844e-07,
      "loss": 0.3777,
      "step": 1521
    },
    {
      "epoch": 2.7,
      "learning_rate": 6.03954463750749e-07,
      "loss": 0.3932,
      "step": 1522
    },
    {
      "epoch": 2.71,
      "learning_rate": 6.003594967046135e-07,
      "loss": 0.404,
      "step": 1523
    },
    {
      "epoch": 2.71,
      "learning_rate": 5.967645296584781e-07,
      "loss": 0.3878,
      "step": 1524
    },
    {
      "epoch": 2.71,
      "learning_rate": 5.931695626123428e-07,
      "loss": 0.356,
      "step": 1525
    },
    {
      "epoch": 2.71,
      "learning_rate": 5.895745955662073e-07,
      "loss": 0.3679,
      "step": 1526
    },
    {
      "epoch": 2.71,
      "learning_rate": 5.859796285200719e-07,
      "loss": 0.3812,
      "step": 1527
    },
    {
      "epoch": 2.71,
      "learning_rate": 5.823846614739365e-07,
      "loss": 0.3858,
      "step": 1528
    },
    {
      "epoch": 2.72,
      "learning_rate": 5.787896944278011e-07,
      "loss": 0.3832,
      "step": 1529
    },
    {
      "epoch": 2.72,
      "learning_rate": 5.751947273816657e-07,
      "loss": 0.3496,
      "step": 1530
    },
    {
      "epoch": 2.72,
      "learning_rate": 5.715997603355302e-07,
      "loss": 0.3881,
      "step": 1531
    },
    {
      "epoch": 2.72,
      "learning_rate": 5.680047932893949e-07,
      "loss": 0.3938,
      "step": 1532
    },
    {
      "epoch": 2.72,
      "learning_rate": 5.644098262432594e-07,
      "loss": 0.3606,
      "step": 1533
    },
    {
      "epoch": 2.72,
      "learning_rate": 5.60814859197124e-07,
      "loss": 0.4045,
      "step": 1534
    },
    {
      "epoch": 2.73,
      "learning_rate": 5.572198921509886e-07,
      "loss": 0.3665,
      "step": 1535
    },
    {
      "epoch": 2.73,
      "learning_rate": 5.536249251048532e-07,
      "loss": 0.3811,
      "step": 1536
    },
    {
      "epoch": 2.73,
      "learning_rate": 5.500299580587178e-07,
      "loss": 0.368,
      "step": 1537
    },
    {
      "epoch": 2.73,
      "learning_rate": 5.464349910125824e-07,
      "loss": 0.3735,
      "step": 1538
    },
    {
      "epoch": 2.73,
      "learning_rate": 5.428400239664471e-07,
      "loss": 0.3576,
      "step": 1539
    },
    {
      "epoch": 2.74,
      "learning_rate": 5.392450569203116e-07,
      "loss": 0.3745,
      "step": 1540
    },
    {
      "epoch": 2.74,
      "learning_rate": 5.356500898741761e-07,
      "loss": 0.3751,
      "step": 1541
    },
    {
      "epoch": 2.74,
      "learning_rate": 5.320551228280408e-07,
      "loss": 0.4006,
      "step": 1542
    },
    {
      "epoch": 2.74,
      "learning_rate": 5.284601557819054e-07,
      "loss": 0.3867,
      "step": 1543
    },
    {
      "epoch": 2.74,
      "learning_rate": 5.2486518873577e-07,
      "loss": 0.3694,
      "step": 1544
    },
    {
      "epoch": 2.74,
      "learning_rate": 5.212702216896345e-07,
      "loss": 0.382,
      "step": 1545
    },
    {
      "epoch": 2.75,
      "learning_rate": 5.17675254643499e-07,
      "loss": 0.3742,
      "step": 1546
    },
    {
      "epoch": 2.75,
      "learning_rate": 5.140802875973637e-07,
      "loss": 0.374,
      "step": 1547
    },
    {
      "epoch": 2.75,
      "learning_rate": 5.104853205512283e-07,
      "loss": 0.376,
      "step": 1548
    },
    {
      "epoch": 2.75,
      "learning_rate": 5.068903535050929e-07,
      "loss": 0.3588,
      "step": 1549
    },
    {
      "epoch": 2.75,
      "learning_rate": 5.032953864589575e-07,
      "loss": 0.3821,
      "step": 1550
    },
    {
      "epoch": 2.75,
      "learning_rate": 4.99700419412822e-07,
      "loss": 0.4012,
      "step": 1551
    },
    {
      "epoch": 2.76,
      "learning_rate": 4.961054523666867e-07,
      "loss": 0.3735,
      "step": 1552
    },
    {
      "epoch": 2.76,
      "learning_rate": 4.925104853205512e-07,
      "loss": 0.3929,
      "step": 1553
    },
    {
      "epoch": 2.76,
      "learning_rate": 4.889155182744159e-07,
      "loss": 0.3576,
      "step": 1554
    },
    {
      "epoch": 2.76,
      "learning_rate": 4.853205512282804e-07,
      "loss": 0.383,
      "step": 1555
    },
    {
      "epoch": 2.76,
      "learning_rate": 4.817255841821449e-07,
      "loss": 0.404,
      "step": 1556
    },
    {
      "epoch": 2.77,
      "learning_rate": 4.781306171360097e-07,
      "loss": 0.3946,
      "step": 1557
    },
    {
      "epoch": 2.77,
      "learning_rate": 4.7453565008987423e-07,
      "loss": 0.3718,
      "step": 1558
    },
    {
      "epoch": 2.77,
      "learning_rate": 4.7094068304373877e-07,
      "loss": 0.3895,
      "step": 1559
    },
    {
      "epoch": 2.77,
      "learning_rate": 4.6734571599760336e-07,
      "loss": 0.3924,
      "step": 1560
    },
    {
      "epoch": 2.77,
      "learning_rate": 4.63750748951468e-07,
      "loss": 0.3806,
      "step": 1561
    },
    {
      "epoch": 2.77,
      "learning_rate": 4.6015578190533254e-07,
      "loss": 0.3618,
      "step": 1562
    },
    {
      "epoch": 2.78,
      "learning_rate": 4.5656081485919713e-07,
      "loss": 0.3918,
      "step": 1563
    },
    {
      "epoch": 2.78,
      "learning_rate": 4.5296584781306167e-07,
      "loss": 0.367,
      "step": 1564
    },
    {
      "epoch": 2.78,
      "learning_rate": 4.4937088076692636e-07,
      "loss": 0.381,
      "step": 1565
    },
    {
      "epoch": 2.78,
      "learning_rate": 4.457759137207909e-07,
      "loss": 0.3848,
      "step": 1566
    },
    {
      "epoch": 2.78,
      "learning_rate": 4.421809466746555e-07,
      "loss": 0.373,
      "step": 1567
    },
    {
      "epoch": 2.79,
      "learning_rate": 4.3858597962852014e-07,
      "loss": 0.3728,
      "step": 1568
    },
    {
      "epoch": 2.79,
      "learning_rate": 4.349910125823847e-07,
      "loss": 0.3906,
      "step": 1569
    },
    {
      "epoch": 2.79,
      "learning_rate": 4.3139604553624926e-07,
      "loss": 0.3942,
      "step": 1570
    },
    {
      "epoch": 2.79,
      "learning_rate": 4.278010784901138e-07,
      "loss": 0.3747,
      "step": 1571
    },
    {
      "epoch": 2.79,
      "learning_rate": 4.2420611144397845e-07,
      "loss": 0.3779,
      "step": 1572
    },
    {
      "epoch": 2.79,
      "learning_rate": 4.2061114439784304e-07,
      "loss": 0.3658,
      "step": 1573
    },
    {
      "epoch": 2.8,
      "learning_rate": 4.1701617735170763e-07,
      "loss": 0.3689,
      "step": 1574
    },
    {
      "epoch": 2.8,
      "learning_rate": 4.1342121030557217e-07,
      "loss": 0.3799,
      "step": 1575
    },
    {
      "epoch": 2.8,
      "learning_rate": 4.098262432594368e-07,
      "loss": 0.3565,
      "step": 1576
    },
    {
      "epoch": 2.8,
      "learning_rate": 4.062312762133014e-07,
      "loss": 0.3823,
      "step": 1577
    },
    {
      "epoch": 2.8,
      "learning_rate": 4.0263630916716594e-07,
      "loss": 0.4153,
      "step": 1578
    },
    {
      "epoch": 2.8,
      "learning_rate": 3.990413421210306e-07,
      "loss": 0.3827,
      "step": 1579
    },
    {
      "epoch": 2.81,
      "learning_rate": 3.9544637507489517e-07,
      "loss": 0.3529,
      "step": 1580
    },
    {
      "epoch": 2.81,
      "learning_rate": 3.9185140802875976e-07,
      "loss": 0.378,
      "step": 1581
    },
    {
      "epoch": 2.81,
      "learning_rate": 3.882564409826243e-07,
      "loss": 0.3787,
      "step": 1582
    },
    {
      "epoch": 2.81,
      "learning_rate": 3.8466147393648894e-07,
      "loss": 0.3647,
      "step": 1583
    },
    {
      "epoch": 2.81,
      "learning_rate": 3.8106650689035354e-07,
      "loss": 0.3693,
      "step": 1584
    },
    {
      "epoch": 2.82,
      "learning_rate": 3.774715398442181e-07,
      "loss": 0.4008,
      "step": 1585
    },
    {
      "epoch": 2.82,
      "learning_rate": 3.738765727980827e-07,
      "loss": 0.3486,
      "step": 1586
    },
    {
      "epoch": 2.82,
      "learning_rate": 3.7028160575194726e-07,
      "loss": 0.3874,
      "step": 1587
    },
    {
      "epoch": 2.82,
      "learning_rate": 3.6668663870581185e-07,
      "loss": 0.3915,
      "step": 1588
    },
    {
      "epoch": 2.82,
      "learning_rate": 3.630916716596765e-07,
      "loss": 0.411,
      "step": 1589
    },
    {
      "epoch": 2.82,
      "learning_rate": 3.5949670461354103e-07,
      "loss": 0.3738,
      "step": 1590
    },
    {
      "epoch": 2.83,
      "learning_rate": 3.5590173756740567e-07,
      "loss": 0.4087,
      "step": 1591
    },
    {
      "epoch": 2.83,
      "learning_rate": 3.523067705212702e-07,
      "loss": 0.401,
      "step": 1592
    },
    {
      "epoch": 2.83,
      "learning_rate": 3.4871180347513485e-07,
      "loss": 0.377,
      "step": 1593
    },
    {
      "epoch": 2.83,
      "learning_rate": 3.451168364289994e-07,
      "loss": 0.3858,
      "step": 1594
    },
    {
      "epoch": 2.83,
      "learning_rate": 3.41521869382864e-07,
      "loss": 0.3789,
      "step": 1595
    },
    {
      "epoch": 2.83,
      "learning_rate": 3.3792690233672857e-07,
      "loss": 0.3937,
      "step": 1596
    },
    {
      "epoch": 2.84,
      "learning_rate": 3.3433193529059316e-07,
      "loss": 0.4194,
      "step": 1597
    },
    {
      "epoch": 2.84,
      "learning_rate": 3.307369682444578e-07,
      "loss": 0.3739,
      "step": 1598
    },
    {
      "epoch": 2.84,
      "learning_rate": 3.2714200119832234e-07,
      "loss": 0.3842,
      "step": 1599
    },
    {
      "epoch": 2.84,
      "learning_rate": 3.23547034152187e-07,
      "loss": 0.4039,
      "step": 1600
    },
    {
      "epoch": 2.84,
      "learning_rate": 3.199520671060515e-07,
      "loss": 0.3839,
      "step": 1601
    },
    {
      "epoch": 2.85,
      "learning_rate": 3.163571000599161e-07,
      "loss": 0.3709,
      "step": 1602
    },
    {
      "epoch": 2.85,
      "learning_rate": 3.127621330137807e-07,
      "loss": 0.4059,
      "step": 1603
    },
    {
      "epoch": 2.85,
      "learning_rate": 3.091671659676453e-07,
      "loss": 0.3762,
      "step": 1604
    },
    {
      "epoch": 2.85,
      "learning_rate": 3.055721989215099e-07,
      "loss": 0.3775,
      "step": 1605
    },
    {
      "epoch": 2.85,
      "learning_rate": 3.019772318753745e-07,
      "loss": 0.382,
      "step": 1606
    },
    {
      "epoch": 2.85,
      "learning_rate": 2.9838226482923907e-07,
      "loss": 0.3826,
      "step": 1607
    },
    {
      "epoch": 2.86,
      "learning_rate": 2.9478729778310366e-07,
      "loss": 0.3742,
      "step": 1608
    },
    {
      "epoch": 2.86,
      "learning_rate": 2.9119233073696825e-07,
      "loss": 0.3682,
      "step": 1609
    },
    {
      "epoch": 2.86,
      "learning_rate": 2.8759736369083284e-07,
      "loss": 0.366,
      "step": 1610
    },
    {
      "epoch": 2.86,
      "learning_rate": 2.8400239664469743e-07,
      "loss": 0.388,
      "step": 1611
    },
    {
      "epoch": 2.86,
      "learning_rate": 2.80407429598562e-07,
      "loss": 0.3833,
      "step": 1612
    },
    {
      "epoch": 2.87,
      "learning_rate": 2.768124625524266e-07,
      "loss": 0.3726,
      "step": 1613
    },
    {
      "epoch": 2.87,
      "learning_rate": 2.732174955062912e-07,
      "loss": 0.3845,
      "step": 1614
    },
    {
      "epoch": 2.87,
      "learning_rate": 2.696225284601558e-07,
      "loss": 0.3998,
      "step": 1615
    },
    {
      "epoch": 2.87,
      "learning_rate": 2.660275614140204e-07,
      "loss": 0.3613,
      "step": 1616
    },
    {
      "epoch": 2.87,
      "learning_rate": 2.62432594367885e-07,
      "loss": 0.3775,
      "step": 1617
    },
    {
      "epoch": 2.87,
      "learning_rate": 2.588376273217495e-07,
      "loss": 0.3851,
      "step": 1618
    },
    {
      "epoch": 2.88,
      "learning_rate": 2.5524266027561416e-07,
      "loss": 0.3801,
      "step": 1619
    },
    {
      "epoch": 2.88,
      "learning_rate": 2.5164769322947875e-07,
      "loss": 0.3662,
      "step": 1620
    },
    {
      "epoch": 2.88,
      "learning_rate": 2.4805272618334334e-07,
      "loss": 0.3879,
      "step": 1621
    },
    {
      "epoch": 2.88,
      "learning_rate": 2.4445775913720793e-07,
      "loss": 0.3524,
      "step": 1622
    },
    {
      "epoch": 2.88,
      "learning_rate": 2.4086279209107247e-07,
      "loss": 0.3745,
      "step": 1623
    },
    {
      "epoch": 2.88,
      "learning_rate": 2.3726782504493711e-07,
      "loss": 0.383,
      "step": 1624
    },
    {
      "epoch": 2.89,
      "learning_rate": 2.3367285799880168e-07,
      "loss": 0.3772,
      "step": 1625
    },
    {
      "epoch": 2.89,
      "learning_rate": 2.3007789095266627e-07,
      "loss": 0.4023,
      "step": 1626
    },
    {
      "epoch": 2.89,
      "learning_rate": 2.2648292390653083e-07,
      "loss": 0.403,
      "step": 1627
    },
    {
      "epoch": 2.89,
      "learning_rate": 2.2288795686039545e-07,
      "loss": 0.377,
      "step": 1628
    },
    {
      "epoch": 2.89,
      "learning_rate": 2.1929298981426007e-07,
      "loss": 0.3639,
      "step": 1629
    },
    {
      "epoch": 2.9,
      "learning_rate": 2.1569802276812463e-07,
      "loss": 0.3633,
      "step": 1630
    },
    {
      "epoch": 2.9,
      "learning_rate": 2.1210305572198922e-07,
      "loss": 0.3836,
      "step": 1631
    },
    {
      "epoch": 2.9,
      "learning_rate": 2.0850808867585381e-07,
      "loss": 0.3753,
      "step": 1632
    },
    {
      "epoch": 2.9,
      "learning_rate": 2.049131216297184e-07,
      "loss": 0.3853,
      "step": 1633
    },
    {
      "epoch": 2.9,
      "learning_rate": 2.0131815458358297e-07,
      "loss": 0.3856,
      "step": 1634
    },
    {
      "epoch": 2.9,
      "learning_rate": 1.9772318753744759e-07,
      "loss": 0.3768,
      "step": 1635
    },
    {
      "epoch": 2.91,
      "learning_rate": 1.9412822049131215e-07,
      "loss": 0.391,
      "step": 1636
    },
    {
      "epoch": 2.91,
      "learning_rate": 1.9053325344517677e-07,
      "loss": 0.3993,
      "step": 1637
    },
    {
      "epoch": 2.91,
      "learning_rate": 1.8693828639904136e-07,
      "loss": 0.3922,
      "step": 1638
    },
    {
      "epoch": 2.91,
      "learning_rate": 1.8334331935290592e-07,
      "loss": 0.3827,
      "step": 1639
    },
    {
      "epoch": 2.91,
      "learning_rate": 1.7974835230677051e-07,
      "loss": 0.3501,
      "step": 1640
    },
    {
      "epoch": 2.91,
      "learning_rate": 1.761533852606351e-07,
      "loss": 0.3802,
      "step": 1641
    },
    {
      "epoch": 2.92,
      "learning_rate": 1.725584182144997e-07,
      "loss": 0.382,
      "step": 1642
    },
    {
      "epoch": 2.92,
      "learning_rate": 1.6896345116836429e-07,
      "loss": 0.397,
      "step": 1643
    },
    {
      "epoch": 2.92,
      "learning_rate": 1.653684841222289e-07,
      "loss": 0.3547,
      "step": 1644
    },
    {
      "epoch": 2.92,
      "learning_rate": 1.617735170760935e-07,
      "loss": 0.4033,
      "step": 1645
    },
    {
      "epoch": 2.92,
      "learning_rate": 1.5817855002995806e-07,
      "loss": 0.3999,
      "step": 1646
    },
    {
      "epoch": 2.93,
      "learning_rate": 1.5458358298382265e-07,
      "loss": 0.3769,
      "step": 1647
    },
    {
      "epoch": 2.93,
      "learning_rate": 1.5098861593768724e-07,
      "loss": 0.379,
      "step": 1648
    },
    {
      "epoch": 2.93,
      "learning_rate": 1.4739364889155183e-07,
      "loss": 0.3906,
      "step": 1649
    },
    {
      "epoch": 2.93,
      "learning_rate": 1.4379868184541642e-07,
      "loss": 0.3913,
      "step": 1650
    },
    {
      "epoch": 2.93,
      "learning_rate": 1.40203714799281e-07,
      "loss": 0.3885,
      "step": 1651
    },
    {
      "epoch": 2.93,
      "learning_rate": 1.366087477531456e-07,
      "loss": 0.3714,
      "step": 1652
    },
    {
      "epoch": 2.94,
      "learning_rate": 1.330137807070102e-07,
      "loss": 0.3876,
      "step": 1653
    },
    {
      "epoch": 2.94,
      "learning_rate": 1.2941881366087476e-07,
      "loss": 0.3796,
      "step": 1654
    },
    {
      "epoch": 2.94,
      "learning_rate": 1.2582384661473938e-07,
      "loss": 0.3756,
      "step": 1655
    },
    {
      "epoch": 2.94,
      "learning_rate": 1.2222887956860397e-07,
      "loss": 0.3915,
      "step": 1656
    },
    {
      "epoch": 2.94,
      "learning_rate": 1.1863391252246856e-07,
      "loss": 0.3719,
      "step": 1657
    },
    {
      "epoch": 2.94,
      "learning_rate": 1.1503894547633313e-07,
      "loss": 0.3815,
      "step": 1658
    },
    {
      "epoch": 2.95,
      "learning_rate": 1.1144397843019773e-07,
      "loss": 0.3925,
      "step": 1659
    },
    {
      "epoch": 2.95,
      "learning_rate": 1.0784901138406232e-07,
      "loss": 0.3526,
      "step": 1660
    },
    {
      "epoch": 2.95,
      "learning_rate": 1.0425404433792691e-07,
      "loss": 0.3902,
      "step": 1661
    },
    {
      "epoch": 2.95,
      "learning_rate": 1.0065907729179148e-07,
      "loss": 0.4026,
      "step": 1662
    },
    {
      "epoch": 2.95,
      "learning_rate": 9.706411024565608e-08,
      "loss": 0.3939,
      "step": 1663
    },
    {
      "epoch": 2.96,
      "learning_rate": 9.346914319952068e-08,
      "loss": 0.3857,
      "step": 1664
    },
    {
      "epoch": 2.96,
      "learning_rate": 8.987417615338526e-08,
      "loss": 0.3996,
      "step": 1665
    },
    {
      "epoch": 2.96,
      "learning_rate": 8.627920910724985e-08,
      "loss": 0.3682,
      "step": 1666
    },
    {
      "epoch": 2.96,
      "learning_rate": 8.268424206111445e-08,
      "loss": 0.3898,
      "step": 1667
    },
    {
      "epoch": 2.96,
      "learning_rate": 7.908927501497903e-08,
      "loss": 0.359,
      "step": 1668
    },
    {
      "epoch": 2.96,
      "learning_rate": 7.549430796884362e-08,
      "loss": 0.3919,
      "step": 1669
    },
    {
      "epoch": 2.97,
      "learning_rate": 7.189934092270821e-08,
      "loss": 0.3933,
      "step": 1670
    },
    {
      "epoch": 2.97,
      "learning_rate": 6.83043738765728e-08,
      "loss": 0.3699,
      "step": 1671
    },
    {
      "epoch": 2.97,
      "learning_rate": 6.470940683043738e-08,
      "loss": 0.4004,
      "step": 1672
    },
    {
      "epoch": 2.97,
      "learning_rate": 6.111443978430198e-08,
      "loss": 0.3964,
      "step": 1673
    },
    {
      "epoch": 2.97,
      "learning_rate": 5.751947273816657e-08,
      "loss": 0.3836,
      "step": 1674
    },
    {
      "epoch": 2.98,
      "learning_rate": 5.392450569203116e-08,
      "loss": 0.4036,
      "step": 1675
    },
    {
      "epoch": 2.98,
      "learning_rate": 5.032953864589574e-08,
      "loss": 0.3688,
      "step": 1676
    },
    {
      "epoch": 2.98,
      "learning_rate": 4.673457159976034e-08,
      "loss": 0.3764,
      "step": 1677
    },
    {
      "epoch": 2.98,
      "learning_rate": 4.3139604553624924e-08,
      "loss": 0.391,
      "step": 1678
    },
    {
      "epoch": 2.98,
      "learning_rate": 3.9544637507489515e-08,
      "loss": 0.3851,
      "step": 1679
    },
    {
      "epoch": 2.98,
      "learning_rate": 3.5949670461354105e-08,
      "loss": 0.3747,
      "step": 1680
    },
    {
      "epoch": 2.99,
      "learning_rate": 3.235470341521869e-08,
      "loss": 0.3838,
      "step": 1681
    },
    {
      "epoch": 2.99,
      "learning_rate": 2.8759736369083284e-08,
      "loss": 0.3735,
      "step": 1682
    },
    {
      "epoch": 2.99,
      "learning_rate": 2.516476932294787e-08,
      "loss": 0.3779,
      "step": 1683
    },
    {
      "epoch": 2.99,
      "learning_rate": 2.1569802276812462e-08,
      "loss": 0.3876,
      "step": 1684
    },
    {
      "epoch": 2.99,
      "learning_rate": 1.7974835230677053e-08,
      "loss": 0.4162,
      "step": 1685
    },
    {
      "epoch": 2.99,
      "learning_rate": 1.4379868184541642e-08,
      "loss": 0.4096,
      "step": 1686
    },
    {
      "epoch": 3.0,
      "learning_rate": 1.0784901138406231e-08,
      "loss": 0.4114,
      "step": 1687
    },
    {
      "epoch": 3.0,
      "learning_rate": 7.189934092270821e-09,
      "loss": 0.377,
      "step": 1688
    },
    {
      "epoch": 3.0,
      "learning_rate": 3.5949670461354105e-09,
      "loss": 0.38,
      "step": 1689
    }
  ],
  "logging_steps": 1,
  "max_steps": 1689,
  "num_train_epochs": 3,
  "save_steps": 500,
  "total_flos": 1.4167476619953832e+19,
  "trial_name": null,
  "trial_params": null
}