{
  "best_metric": null,
  "best_model_checkpoint": null,
  "epoch": 0.3773940937824323,
  "global_step": 1000,
  "is_hyper_param_search": false,
  "is_local_process_zero": true,
  "is_world_process_zero": true,
  "log_history": [
    {
      "epoch": 0.0,
      "learning_rate": 4.000000000000001e-06,
      "loss": 1.3771,
      "step": 1
    },
    {
      "epoch": 0.0,
      "learning_rate": 8.000000000000001e-06,
      "loss": 1.6683,
      "step": 2
    },
    {
      "epoch": 0.0,
      "learning_rate": 1.2e-05,
      "loss": 1.6333,
      "step": 3
    },
    {
      "epoch": 0.0,
      "learning_rate": 1.6000000000000003e-05,
      "loss": 1.4646,
      "step": 4
    },
    {
      "epoch": 0.0,
      "learning_rate": 2e-05,
      "loss": 1.4043,
      "step": 5
    },
    {
      "epoch": 0.0,
      "learning_rate": 2.4e-05,
      "loss": 1.6339,
      "step": 6
    },
    {
      "epoch": 0.0,
      "learning_rate": 2.8000000000000003e-05,
      "loss": 1.8558,
      "step": 7
    },
    {
      "epoch": 0.0,
      "learning_rate": 3.2000000000000005e-05,
      "loss": 1.4294,
      "step": 8
    },
    {
      "epoch": 0.0,
      "learning_rate": 3.6e-05,
      "loss": 1.5738,
      "step": 9
    },
    {
      "epoch": 0.0,
      "learning_rate": 4e-05,
      "loss": 1.5463,
      "step": 10
    },
    {
      "epoch": 0.0,
      "learning_rate": 4.4000000000000006e-05,
      "loss": 1.6751,
      "step": 11
    },
    {
      "epoch": 0.0,
      "learning_rate": 4.8e-05,
      "loss": 1.3243,
      "step": 12
    },
    {
      "epoch": 0.0,
      "learning_rate": 5.2000000000000004e-05,
      "loss": 1.6727,
      "step": 13
    },
    {
      "epoch": 0.01,
      "learning_rate": 5.6000000000000006e-05,
      "loss": 1.4421,
      "step": 14
    },
    {
      "epoch": 0.01,
      "learning_rate": 6e-05,
      "loss": 1.5361,
      "step": 15
    },
    {
      "epoch": 0.01,
      "learning_rate": 6.400000000000001e-05,
      "loss": 1.5129,
      "step": 16
    },
    {
      "epoch": 0.01,
      "learning_rate": 6.800000000000001e-05,
      "loss": 1.3696,
      "step": 17
    },
    {
      "epoch": 0.01,
      "learning_rate": 7.2e-05,
      "loss": 1.4057,
      "step": 18
    },
    {
      "epoch": 0.01,
      "learning_rate": 7.6e-05,
      "loss": 1.473,
      "step": 19
    },
    {
      "epoch": 0.01,
      "learning_rate": 8e-05,
      "loss": 1.5723,
      "step": 20
    },
    {
      "epoch": 0.01,
      "learning_rate": 8.4e-05,
      "loss": 1.5101,
      "step": 21
    },
    {
      "epoch": 0.01,
      "learning_rate": 8.800000000000001e-05,
      "loss": 1.6298,
      "step": 22
    },
    {
      "epoch": 0.01,
      "learning_rate": 9.200000000000001e-05,
      "loss": 1.4543,
      "step": 23
    },
    {
      "epoch": 0.01,
      "learning_rate": 9.6e-05,
      "loss": 1.4296,
      "step": 24
    },
    {
      "epoch": 0.01,
      "learning_rate": 0.0001,
      "loss": 1.529,
      "step": 25
    },
    {
      "epoch": 0.01,
      "learning_rate": 0.00010400000000000001,
      "loss": 1.6653,
      "step": 26
    },
    {
      "epoch": 0.01,
      "learning_rate": 0.00010800000000000001,
      "loss": 1.6178,
      "step": 27
    },
    {
      "epoch": 0.01,
      "learning_rate": 0.00011200000000000001,
      "loss": 1.4393,
      "step": 28
    },
    {
      "epoch": 0.01,
      "learning_rate": 0.000116,
      "loss": 1.4561,
      "step": 29
    },
    {
      "epoch": 0.01,
      "learning_rate": 0.00012,
      "loss": 1.5368,
      "step": 30
    },
    {
      "epoch": 0.01,
      "learning_rate": 0.000124,
      "loss": 1.5757,
      "step": 31
    },
    {
      "epoch": 0.01,
      "learning_rate": 0.00012800000000000002,
      "loss": 1.4725,
      "step": 32
    },
    {
      "epoch": 0.01,
      "learning_rate": 0.000132,
      "loss": 1.6976,
      "step": 33
    },
    {
      "epoch": 0.01,
      "learning_rate": 0.00013600000000000003,
      "loss": 1.4537,
      "step": 34
    },
    {
      "epoch": 0.01,
      "learning_rate": 0.00014,
      "loss": 1.3886,
      "step": 35
    },
    {
      "epoch": 0.01,
      "learning_rate": 0.000144,
      "loss": 1.4515,
      "step": 36
    },
    {
      "epoch": 0.01,
      "learning_rate": 0.000148,
      "loss": 1.3569,
      "step": 37
    },
    {
      "epoch": 0.01,
      "learning_rate": 0.000152,
      "loss": 1.3558,
      "step": 38
    },
    {
      "epoch": 0.01,
      "learning_rate": 0.00015600000000000002,
      "loss": 1.5902,
      "step": 39
    },
    {
      "epoch": 0.02,
      "learning_rate": 0.00016,
      "loss": 1.3728,
      "step": 40
    },
    {
      "epoch": 0.02,
      "learning_rate": 0.000164,
      "loss": 1.4576,
      "step": 41
    },
    {
      "epoch": 0.02,
      "learning_rate": 0.000168,
      "loss": 1.3009,
      "step": 42
    },
    {
      "epoch": 0.02,
      "learning_rate": 0.000172,
      "loss": 1.4652,
      "step": 43
    },
    {
      "epoch": 0.02,
      "learning_rate": 0.00017600000000000002,
      "loss": 1.4355,
      "step": 44
    },
    {
      "epoch": 0.02,
      "learning_rate": 0.00018,
      "loss": 1.6715,
      "step": 45
    },
    {
      "epoch": 0.02,
      "learning_rate": 0.00018400000000000003,
      "loss": 1.3025,
      "step": 46
    },
    {
      "epoch": 0.02,
      "learning_rate": 0.000188,
      "loss": 1.3885,
      "step": 47
    },
    {
      "epoch": 0.02,
      "learning_rate": 0.000192,
      "loss": 1.5249,
      "step": 48
    },
    {
      "epoch": 0.02,
      "learning_rate": 0.000196,
      "loss": 1.3855,
      "step": 49
    },
    {
      "epoch": 0.02,
      "learning_rate": 0.0002,
      "loss": 1.7058,
      "step": 50
    },
    {
      "epoch": 0.02,
      "learning_rate": 0.00019999945320801072,
      "loss": 1.4766,
      "step": 51
    },
    {
      "epoch": 0.02,
      "learning_rate": 0.0001999978128380225,
      "loss": 1.4239,
      "step": 52
    },
    {
      "epoch": 0.02,
      "learning_rate": 0.00019999507890797408,
      "loss": 1.527,
      "step": 53
    },
    {
      "epoch": 0.02,
      "learning_rate": 0.0001999912514477634,
      "loss": 1.4989,
      "step": 54
    },
    {
      "epoch": 0.02,
      "learning_rate": 0.0001999863304992469,
      "loss": 1.4985,
      "step": 55
    },
    {
      "epoch": 0.02,
      "learning_rate": 0.0001999803161162393,
      "loss": 1.4964,
      "step": 56
    },
    {
      "epoch": 0.02,
      "learning_rate": 0.0001999732083645129,
      "loss": 1.3822,
      "step": 57
    },
    {
      "epoch": 0.02,
      "learning_rate": 0.00019996500732179695,
      "loss": 1.4295,
      "step": 58
    },
    {
      "epoch": 0.02,
      "learning_rate": 0.0001999557130777767,
      "loss": 1.7715,
      "step": 59
    },
    {
      "epoch": 0.02,
      "learning_rate": 0.00019994532573409262,
      "loss": 1.3596,
      "step": 60
    },
    {
      "epoch": 0.02,
      "learning_rate": 0.00019993384540433894,
      "loss": 1.5048,
      "step": 61
    },
    {
      "epoch": 0.02,
      "learning_rate": 0.00019992127221406275,
      "loss": 1.2653,
      "step": 62
    },
    {
      "epoch": 0.02,
      "learning_rate": 0.00019990760630076237,
      "loss": 1.3854,
      "step": 63
    },
    {
      "epoch": 0.02,
      "learning_rate": 0.00019989284781388617,
      "loss": 1.4032,
      "step": 64
    },
    {
      "epoch": 0.02,
      "learning_rate": 0.00019987699691483048,
      "loss": 1.4055,
      "step": 65
    },
    {
      "epoch": 0.02,
      "learning_rate": 0.00019986005377693825,
      "loss": 1.4952,
      "step": 66
    },
    {
      "epoch": 0.03,
      "learning_rate": 0.00019984201858549693,
      "loss": 1.5547,
      "step": 67
    },
    {
      "epoch": 0.03,
      "learning_rate": 0.00019982289153773646,
      "loss": 1.5564,
      "step": 68
    },
    {
      "epoch": 0.03,
      "learning_rate": 0.00019980267284282717,
      "loss": 1.5104,
      "step": 69
    },
    {
      "epoch": 0.03,
      "learning_rate": 0.00019978136272187747,
      "loss": 1.4518,
      "step": 70
    },
    {
      "epoch": 0.03,
      "learning_rate": 0.00019975896140793142,
      "loss": 1.5248,
      "step": 71
    },
    {
      "epoch": 0.03,
      "learning_rate": 0.00019973546914596623,
      "loss": 1.3548,
      "step": 72
    },
    {
      "epoch": 0.03,
      "learning_rate": 0.0001997108861928895,
      "loss": 1.4315,
      "step": 73
    },
    {
      "epoch": 0.03,
      "learning_rate": 0.00019968521281753642,
      "loss": 1.8023,
      "step": 74
    },
    {
      "epoch": 0.03,
      "learning_rate": 0.000199658449300667,
      "loss": 1.8757,
      "step": 75
    },
    {
      "epoch": 0.03,
      "learning_rate": 0.00019963059593496268,
      "loss": 1.55,
      "step": 76
    },
    {
      "epoch": 0.03,
      "learning_rate": 0.0001996016530250235,
      "loss": 1.9263,
      "step": 77
    },
    {
      "epoch": 0.03,
      "learning_rate": 0.0001995716208873644,
      "loss": 1.5269,
      "step": 78
    },
    {
      "epoch": 0.03,
      "learning_rate": 0.00019954049985041204,
      "loss": 1.6926,
      "step": 79
    },
    {
      "epoch": 0.03,
      "learning_rate": 0.00019950829025450114,
      "loss": 1.1818,
      "step": 80
    },
    {
      "epoch": 0.03,
      "learning_rate": 0.00019947499245187068,
      "loss": 1.2472,
      "step": 81
    },
    {
      "epoch": 0.03,
      "learning_rate": 0.00019944060680666002,
      "loss": 1.6255,
      "step": 82
    },
    {
      "epoch": 0.03,
      "learning_rate": 0.00019940513369490516,
      "loss": 1.5712,
      "step": 83
    },
    {
      "epoch": 0.03,
      "learning_rate": 0.0001993685735045343,
      "loss": 1.3684,
      "step": 84
    },
    {
      "epoch": 0.03,
      "learning_rate": 0.00019933092663536382,
      "loss": 1.462,
      "step": 85
    },
    {
      "epoch": 0.03,
      "learning_rate": 0.00019929219349909392,
      "loss": 1.5298,
      "step": 86
    },
    {
      "epoch": 0.03,
      "learning_rate": 0.0001992523745193039,
      "loss": 1.331,
      "step": 87
    },
    {
      "epoch": 0.03,
      "learning_rate": 0.0001992114701314478,
      "loss": 1.7856,
      "step": 88
    },
    {
      "epoch": 0.03,
      "learning_rate": 0.0001991694807828494,
      "loss": 1.5808,
      "step": 89
    },
    {
      "epoch": 0.03,
      "learning_rate": 0.00019912640693269752,
      "loss": 1.4136,
      "step": 90
    },
    {
      "epoch": 0.03,
      "learning_rate": 0.0001990822490520409,
      "loss": 1.4673,
      "step": 91
    },
    {
      "epoch": 0.03,
      "learning_rate": 0.000199037007623783,
      "loss": 1.6898,
      "step": 92
    },
    {
      "epoch": 0.04,
      "learning_rate": 0.00019899068314267688,
      "loss": 1.6261,
      "step": 93
    },
    {
      "epoch": 0.04,
      "learning_rate": 0.0001989432761153196,
      "loss": 1.4805,
      "step": 94
    },
    {
      "epoch": 0.04,
      "learning_rate": 0.00019889478706014687,
      "loss": 1.5018,
      "step": 95
    },
    {
      "epoch": 0.04,
      "learning_rate": 0.00019884521650742715,
      "loss": 1.6089,
      "step": 96
    },
    {
      "epoch": 0.04,
      "learning_rate": 0.00019879456499925614,
      "loss": 1.6487,
      "step": 97
    },
    {
      "epoch": 0.04,
      "learning_rate": 0.00019874283308955057,
      "loss": 1.517,
      "step": 98
    },
    {
      "epoch": 0.04,
      "learning_rate": 0.00019869002134404235,
      "loss": 1.5386,
      "step": 99
    },
    {
      "epoch": 0.04,
      "learning_rate": 0.00019863613034027224,
      "loss": 1.4188,
      "step": 100
    },
    {
      "epoch": 0.04,
      "learning_rate": 0.00019858116066758362,
      "loss": 1.4651,
      "step": 101
    },
    {
      "epoch": 0.04,
      "learning_rate": 0.00019852511292711608,
      "loss": 1.3594,
      "step": 102
    },
    {
      "epoch": 0.04,
      "learning_rate": 0.00019846798773179866,
      "loss": 1.4322,
      "step": 103
    },
    {
      "epoch": 0.04,
      "learning_rate": 0.0001984097857063434,
      "loss": 1.4874,
      "step": 104
    },
    {
      "epoch": 0.04,
      "learning_rate": 0.00019835050748723824,
      "loss": 1.3447,
      "step": 105
    },
    {
      "epoch": 0.04,
      "learning_rate": 0.00019829015372274038,
      "loss": 1.4409,
      "step": 106
    },
    {
      "epoch": 0.04,
      "learning_rate": 0.0001982287250728689,
      "loss": 1.4706,
      "step": 107
    },
    {
      "epoch": 0.04,
      "learning_rate": 0.0001981662222093976,
      "loss": 1.2554,
      "step": 108
    },
    {
      "epoch": 0.04,
      "learning_rate": 0.00019810264581584787,
      "loss": 1.4362,
      "step": 109
    },
    {
      "epoch": 0.04,
      "learning_rate": 0.00019803799658748094,
      "loss": 1.5449,
      "step": 110
    },
    {
      "epoch": 0.04,
      "learning_rate": 0.0001979722752312904,
      "loss": 1.5537,
      "step": 111
    },
    {
      "epoch": 0.04,
      "learning_rate": 0.00019790548246599447,
      "loss": 1.77,
      "step": 112
    },
    {
      "epoch": 0.04,
      "learning_rate": 0.00019783761902202813,
      "loss": 1.474,
      "step": 113
    },
    {
      "epoch": 0.04,
      "learning_rate": 0.00019776868564153516,
      "loss": 1.5617,
      "step": 114
    },
    {
      "epoch": 0.04,
      "learning_rate": 0.00019769868307835994,
      "loss": 1.1664,
      "step": 115
    },
    {
      "epoch": 0.04,
      "learning_rate": 0.00019762761209803927,
      "loss": 1.2867,
      "step": 116
    },
    {
      "epoch": 0.04,
      "learning_rate": 0.00019755547347779403,
      "loss": 1.5673,
      "step": 117
    },
    {
      "epoch": 0.04,
      "learning_rate": 0.0001974822680065206,
      "loss": 1.5614,
      "step": 118
    },
    {
      "epoch": 0.04,
      "learning_rate": 0.00019740799648478233,
      "loss": 1.5865,
      "step": 119
    },
    {
      "epoch": 0.05,
      "learning_rate": 0.0001973326597248006,
      "loss": 1.4008,
      "step": 120
    },
    {
      "epoch": 0.05,
      "learning_rate": 0.00019725625855044617,
      "loss": 1.7168,
      "step": 121
    },
    {
      "epoch": 0.05,
      "learning_rate": 0.00019717879379723012,
      "loss": 1.4881,
      "step": 122
    },
    {
      "epoch": 0.05,
      "learning_rate": 0.0001971002663122945,
      "loss": 1.7234,
      "step": 123
    },
    {
      "epoch": 0.05,
      "learning_rate": 0.00019702067695440332,
      "loss": 1.5824,
      "step": 124
    },
    {
      "epoch": 0.05,
      "learning_rate": 0.00019694002659393305,
      "loss": 1.4202,
      "step": 125
    },
    {
      "epoch": 0.05,
      "learning_rate": 0.0001968583161128631,
      "loss": 1.3443,
      "step": 126
    },
    {
      "epoch": 0.05,
      "learning_rate": 0.00019677554640476624,
      "loss": 1.4996,
      "step": 127
    },
    {
      "epoch": 0.05,
      "learning_rate": 0.00019669171837479873,
      "loss": 1.6795,
      "step": 128
    },
    {
      "epoch": 0.05,
      "learning_rate": 0.00019660683293969041,
      "loss": 1.4691,
      "step": 129
    },
    {
      "epoch": 0.05,
      "learning_rate": 0.00019652089102773488,
      "loss": 1.6054,
      "step": 130
    },
    {
      "epoch": 0.05,
      "learning_rate": 0.00019643389357877907,
      "loss": 1.5673,
      "step": 131
    },
    {
      "epoch": 0.05,
      "learning_rate": 0.00019634584154421317,
      "loss": 1.4519,
      "step": 132
    },
    {
      "epoch": 0.05,
      "learning_rate": 0.00019625673588696008,
      "loss": 1.6135,
      "step": 133
    },
    {
      "epoch": 0.05,
      "learning_rate": 0.00019616657758146503,
      "loss": 1.4928,
      "step": 134
    },
    {
      "epoch": 0.05,
      "learning_rate": 0.00019607536761368484,
      "loss": 1.4667,
      "step": 135
    },
    {
      "epoch": 0.05,
      "learning_rate": 0.00019598310698107702,
      "loss": 1.5076,
      "step": 136
    },
    {
      "epoch": 0.05,
      "learning_rate": 0.0001958897966925891,
      "loss": 1.6252,
      "step": 137
    },
    {
      "epoch": 0.05,
      "learning_rate": 0.0001957954377686475,
      "loss": 1.5067,
      "step": 138
    },
    {
      "epoch": 0.05,
      "learning_rate": 0.00019570003124114619,
      "loss": 1.4797,
      "step": 139
    },
    {
      "epoch": 0.05,
      "learning_rate": 0.00019560357815343577,
      "loss": 1.5738,
      "step": 140
    },
    {
      "epoch": 0.05,
      "learning_rate": 0.0001955060795603117,
      "loss": 1.5238,
      "step": 141
    },
    {
      "epoch": 0.05,
      "learning_rate": 0.000195407536528003,
      "loss": 1.713,
      "step": 142
    },
    {
      "epoch": 0.05,
      "learning_rate": 0.00019530795013416046,
      "loss": 1.6015,
      "step": 143
    },
    {
      "epoch": 0.05,
      "learning_rate": 0.00019520732146784491,
      "loss": 1.6067,
      "step": 144
    },
    {
      "epoch": 0.05,
      "learning_rate": 0.00019510565162951537,
      "loss": 1.6063,
      "step": 145
    },
    {
      "epoch": 0.06,
      "learning_rate": 0.00019500294173101687,
      "loss": 1.6428,
      "step": 146
    },
    {
      "epoch": 0.06,
      "learning_rate": 0.00019489919289556845,
      "loss": 1.4835,
      "step": 147
    },
    {
      "epoch": 0.06,
      "learning_rate": 0.0001947944062577507,
      "loss": 1.491,
      "step": 148
    },
    {
      "epoch": 0.06,
      "learning_rate": 0.0001946885829634935,
      "loss": 1.4183,
      "step": 149
    },
    {
      "epoch": 0.06,
      "learning_rate": 0.00019458172417006347,
      "loss": 1.4366,
      "step": 150
    },
    {
      "epoch": 0.06,
      "learning_rate": 0.00019447383104605125,
      "loss": 1.5319,
      "step": 151
    },
    {
      "epoch": 0.06,
      "learning_rate": 0.00019436490477135878,
      "loss": 1.4926,
      "step": 152
    },
    {
      "epoch": 0.06,
      "learning_rate": 0.0001942549465371863,
      "loss": 1.7489,
      "step": 153
    },
    {
      "epoch": 0.06,
      "learning_rate": 0.00019414395754601947,
      "loss": 1.4453,
      "step": 154
    },
    {
      "epoch": 0.06,
      "learning_rate": 0.00019403193901161613,
      "loss": 1.4814,
      "step": 155
    },
    {
      "epoch": 0.06,
      "learning_rate": 0.00019391889215899299,
      "loss": 1.6483,
      "step": 156
    },
    {
      "epoch": 0.06,
      "learning_rate": 0.00019380481822441235,
      "loss": 1.5119,
      "step": 157
    },
    {
      "epoch": 0.06,
      "learning_rate": 0.00019368971845536845,
      "loss": 1.318,
      "step": 158
    },
    {
      "epoch": 0.06,
      "learning_rate": 0.000193573594110574,
      "loss": 1.4537,
      "step": 159
    },
    {
      "epoch": 0.06,
      "learning_rate": 0.0001934564464599461,
      "loss": 1.5204,
      "step": 160
    },
    {
      "epoch": 0.06,
      "learning_rate": 0.0001933382767845928,
      "loss": 1.4601,
      "step": 161
    },
    {
      "epoch": 0.06,
      "learning_rate": 0.00019321908637679865,
      "loss": 1.5626,
      "step": 162
    },
    {
      "epoch": 0.06,
      "learning_rate": 0.00019309887654001096,
      "loss": 1.4477,
      "step": 163
    },
    {
      "epoch": 0.06,
      "learning_rate": 0.00019297764858882514,
      "loss": 1.5851,
      "step": 164
    },
    {
      "epoch": 0.06,
      "learning_rate": 0.00019285540384897073,
      "loss": 1.555,
      "step": 165
    },
    {
      "epoch": 0.06,
      "learning_rate": 0.00019273214365729655,
      "loss": 1.4885,
      "step": 166
    },
    {
      "epoch": 0.06,
      "learning_rate": 0.00019260786936175635,
      "loss": 1.6686,
      "step": 167
    },
    {
      "epoch": 0.06,
      "learning_rate": 0.00019248258232139388,
      "loss": 1.7091,
      "step": 168
    },
    {
      "epoch": 0.06,
      "learning_rate": 0.00019235628390632822,
      "loss": 1.549,
      "step": 169
    },
    {
      "epoch": 0.06,
      "learning_rate": 0.00019222897549773848,
      "loss": 1.5747,
      "step": 170
    },
    {
      "epoch": 0.06,
      "learning_rate": 0.00019210065848784913,
      "loss": 1.8573,
      "step": 171
    },
    {
      "epoch": 0.06,
      "learning_rate": 0.00019197133427991436,
      "loss": 1.5019,
      "step": 172
    },
    {
      "epoch": 0.07,
      "learning_rate": 0.000191841004288203,
      "loss": 1.5118,
      "step": 173
    },
    {
      "epoch": 0.07,
      "learning_rate": 0.000191709669937983,
      "loss": 1.5818,
      "step": 174
    },
    {
      "epoch": 0.07,
      "learning_rate": 0.00019157733266550575,
      "loss": 1.4686,
      "step": 175
    },
    {
      "epoch": 0.07,
      "learning_rate": 0.00019144399391799043,
      "loss": 1.3446,
      "step": 176
    },
    {
      "epoch": 0.07,
      "learning_rate": 0.0001913096551536083,
      "loss": 1.5599,
      "step": 177
    },
    {
      "epoch": 0.07,
      "learning_rate": 0.00019117431784146645,
      "loss": 1.3847,
      "step": 178
    },
    {
      "epoch": 0.07,
      "learning_rate": 0.00019103798346159213,
      "loss": 1.3977,
      "step": 179
    },
    {
      "epoch": 0.07,
      "learning_rate": 0.00019090065350491626,
      "loss": 1.4674,
      "step": 180
    },
    {
      "epoch": 0.07,
      "learning_rate": 0.00019076232947325722,
      "loss": 1.8562,
      "step": 181
    },
    {
      "epoch": 0.07,
      "learning_rate": 0.00019062301287930446,
      "loss": 1.4907,
      "step": 182
    },
    {
      "epoch": 0.07,
      "learning_rate": 0.00019048270524660196,
      "loss": 1.5626,
      "step": 183
    },
    {
      "epoch": 0.07,
      "learning_rate": 0.0001903414081095315,
      "loss": 1.8103,
      "step": 184
    },
    {
      "epoch": 0.07,
      "learning_rate": 0.00019019912301329592,
      "loss": 1.5239,
      "step": 185
    },
    {
      "epoch": 0.07,
      "learning_rate": 0.00019005585151390223,
      "loss": 1.8052,
      "step": 186
    },
    {
      "epoch": 0.07,
      "learning_rate": 0.0001899115951781446,
      "loss": 1.4232,
      "step": 187
    },
    {
      "epoch": 0.07,
      "learning_rate": 0.00018976635558358722,
      "loss": 1.6923,
      "step": 188
    },
    {
      "epoch": 0.07,
      "learning_rate": 0.00018962013431854702,
      "loss": 1.3907,
      "step": 189
    },
    {
      "epoch": 0.07,
      "learning_rate": 0.00018947293298207635,
      "loss": 1.5926,
      "step": 190
    },
    {
      "epoch": 0.07,
      "learning_rate": 0.0001893247531839454,
      "loss": 1.4904,
      "step": 191
    },
    {
      "epoch": 0.07,
      "learning_rate": 0.00018917559654462474,
      "loss": 1.3902,
      "step": 192
    },
    {
      "epoch": 0.07,
      "learning_rate": 0.00018902546469526743,
      "loss": 1.8517,
      "step": 193
    },
    {
      "epoch": 0.07,
      "learning_rate": 0.00018887435927769137,
      "loss": 1.5293,
      "step": 194
    },
    {
      "epoch": 0.07,
      "learning_rate": 0.0001887222819443612,
      "loss": 1.4155,
      "step": 195
    },
    {
      "epoch": 0.07,
      "learning_rate": 0.00018856923435837022,
      "loss": 1.4577,
      "step": 196
    },
    {
      "epoch": 0.07,
      "learning_rate": 0.00018841521819342236,
      "loss": 1.5775,
      "step": 197
    },
    {
      "epoch": 0.07,
      "learning_rate": 0.0001882602351338137,
      "loss": 1.4697,
      "step": 198
    },
    {
      "epoch": 0.08,
      "learning_rate": 0.00018810428687441414,
      "loss": 1.69,
      "step": 199
    },
    {
      "epoch": 0.08,
      "learning_rate": 0.0001879473751206489,
      "loss": 1.3792,
      "step": 200
    },
    {
      "epoch": 0.08,
      "learning_rate": 0.00018778950158847976,
      "loss": 1.3913,
      "step": 201
    },
    {
      "epoch": 0.08,
      "learning_rate": 0.00018763066800438636,
      "loss": 1.677,
      "step": 202
    },
    {
      "epoch": 0.08,
      "learning_rate": 0.00018747087610534736,
      "loss": 1.6473,
      "step": 203
    },
    {
      "epoch": 0.08,
      "learning_rate": 0.00018731012763882133,
      "loss": 1.5274,
      "step": 204
    },
    {
      "epoch": 0.08,
      "learning_rate": 0.00018714842436272773,
      "loss": 1.6764,
      "step": 205
    },
    {
      "epoch": 0.08,
      "learning_rate": 0.00018698576804542777,
      "loss": 1.4051,
      "step": 206
    },
    {
      "epoch": 0.08,
      "learning_rate": 0.00018682216046570475,
      "loss": 1.4622,
      "step": 207
    },
    {
      "epoch": 0.08,
      "learning_rate": 0.00018665760341274505,
      "loss": 1.7445,
      "step": 208
    },
    {
      "epoch": 0.08,
      "learning_rate": 0.0001864920986861182,
      "loss": 1.6491,
      "step": 209
    },
    {
      "epoch": 0.08,
      "learning_rate": 0.00018632564809575742,
      "loss": 1.8047,
      "step": 210
    },
    {
      "epoch": 0.08,
      "learning_rate": 0.0001861582534619396,
      "loss": 1.6011,
      "step": 211
    },
    {
      "epoch": 0.08,
      "learning_rate": 0.00018598991661526572,
      "loss": 1.4319,
      "step": 212
    },
    {
      "epoch": 0.08,
      "learning_rate": 0.0001858206393966405,
      "loss": 1.6013,
      "step": 213
    },
    {
      "epoch": 0.08,
      "learning_rate": 0.00018565042365725258,
      "loss": 1.5379,
      "step": 214
    },
    {
      "epoch": 0.08,
      "learning_rate": 0.0001854792712585539,
      "loss": 1.5043,
      "step": 215
    },
    {
      "epoch": 0.08,
      "learning_rate": 0.00018530718407223974,
      "loss": 1.8997,
      "step": 216
    },
    {
      "epoch": 0.08,
      "learning_rate": 0.00018513416398022802,
      "loss": 1.6282,
      "step": 217
    },
    {
      "epoch": 0.08,
      "learning_rate": 0.0001849602128746387,
      "loss": 1.6524,
      "step": 218
    },
    {
      "epoch": 0.08,
      "learning_rate": 0.00018478533265777318,
      "loss": 1.6837,
      "step": 219
    },
    {
      "epoch": 0.08,
      "learning_rate": 0.00018460952524209355,
      "loss": 1.3932,
      "step": 220
    },
    {
      "epoch": 0.08,
      "learning_rate": 0.00018443279255020152,
      "loss": 1.5441,
      "step": 221
    },
    {
      "epoch": 0.08,
      "learning_rate": 0.00018425513651481747,
      "loss": 1.4939,
      "step": 222
    },
    {
      "epoch": 0.08,
      "learning_rate": 0.0001840765590787594,
      "loss": 1.6861,
      "step": 223
    },
    {
      "epoch": 0.08,
      "learning_rate": 0.00018389706219492147,
      "loss": 1.6589,
      "step": 224
    },
    {
      "epoch": 0.08,
      "learning_rate": 0.00018371664782625287,
      "loss": 1.4755,
      "step": 225
    },
    {
      "epoch": 0.09,
      "learning_rate": 0.00018353531794573625,
      "loss": 1.5676,
      "step": 226
    },
    {
      "epoch": 0.09,
      "learning_rate": 0.0001833530745363661,
      "loss": 1.6569,
      "step": 227
    },
    {
      "epoch": 0.09,
      "learning_rate": 0.00018316991959112716,
      "loss": 1.6602,
      "step": 228
    },
    {
      "epoch": 0.09,
      "learning_rate": 0.0001829858551129726,
      "loss": 1.5835,
      "step": 229
    },
    {
      "epoch": 0.09,
      "learning_rate": 0.00018280088311480201,
      "loss": 1.4652,
      "step": 230
    },
    {
      "epoch": 0.09,
      "learning_rate": 0.00018261500561943955,
      "loss": 1.3955,
      "step": 231
    },
    {
      "epoch": 0.09,
      "learning_rate": 0.00018242822465961176,
      "loss": 1.7412,
      "step": 232
    },
    {
      "epoch": 0.09,
      "learning_rate": 0.00018224054227792524,
      "loss": 1.6381,
      "step": 233
    },
    {
      "epoch": 0.09,
      "learning_rate": 0.00018205196052684445,
      "loss": 1.6402,
      "step": 234
    },
    {
      "epoch": 0.09,
      "learning_rate": 0.00018186248146866927,
      "loss": 1.6209,
      "step": 235
    },
    {
      "epoch": 0.09,
      "learning_rate": 0.00018167210717551224,
      "loss": 1.5313,
      "step": 236
    },
    {
      "epoch": 0.09,
      "learning_rate": 0.00018148083972927616,
      "loss": 1.7225,
      "step": 237
    },
    {
      "epoch": 0.09,
      "learning_rate": 0.00018128868122163123,
      "loss": 1.615,
      "step": 238
    },
    {
      "epoch": 0.09,
      "learning_rate": 0.000181095633753992,
      "loss": 1.3971,
      "step": 239
    },
    {
      "epoch": 0.09,
      "learning_rate": 0.00018090169943749476,
      "loss": 1.6964,
      "step": 240
    },
    {
      "epoch": 0.09,
      "learning_rate": 0.00018070688039297403,
      "loss": 1.6681,
      "step": 241
    },
    {
      "epoch": 0.09,
      "learning_rate": 0.00018051117875093976,
      "loss": 1.674,
      "step": 242
    },
    {
      "epoch": 0.09,
      "learning_rate": 0.00018031459665155363,
      "loss": 1.3495,
      "step": 243
    },
    {
      "epoch": 0.09,
      "learning_rate": 0.00018011713624460608,
      "loss": 1.4248,
      "step": 244
    },
    {
      "epoch": 0.09,
      "learning_rate": 0.0001799187996894925,
      "loss": 1.7693,
      "step": 245
    },
    {
      "epoch": 0.09,
      "learning_rate": 0.0001797195891551896,
      "loss": 1.7993,
      "step": 246
    },
    {
      "epoch": 0.09,
      "learning_rate": 0.00017951950682023191,
      "loss": 1.5703,
      "step": 247
    },
    {
      "epoch": 0.09,
      "learning_rate": 0.00017931855487268782,
      "loss": 1.5827,
      "step": 248
    },
    {
      "epoch": 0.09,
      "learning_rate": 0.00017911673551013551,
      "loss": 1.4503,
      "step": 249
    },
    {
      "epoch": 0.09,
      "learning_rate": 0.00017891405093963938,
      "loss": 1.6982,
      "step": 250
    },
    {
      "epoch": 0.09,
      "learning_rate": 0.00017871050337772525,
      "loss": 1.5524,
      "step": 251
    },
    {
      "epoch": 0.1,
      "learning_rate": 0.0001785060950503568,
      "loss": 1.6975,
      "step": 252
    },
    {
      "epoch": 0.1,
      "learning_rate": 0.0001783008281929106,
      "loss": 1.6971,
      "step": 253
    },
    {
      "epoch": 0.1,
      "learning_rate": 0.0001780947050501522,
      "loss": 1.7433,
      "step": 254
    },
    {
      "epoch": 0.1,
      "learning_rate": 0.00017788772787621126,
      "loss": 1.587,
      "step": 255
    },
    {
      "epoch": 0.1,
      "learning_rate": 0.00017767989893455698,
      "loss": 1.4709,
      "step": 256
    },
    {
      "epoch": 0.1,
      "learning_rate": 0.00017747122049797335,
      "loss": 1.4217,
      "step": 257
    },
    {
      "epoch": 0.1,
      "learning_rate": 0.00017726169484853438,
      "loss": 1.7229,
      "step": 258
    },
    {
      "epoch": 0.1,
      "learning_rate": 0.00017705132427757895,
      "loss": 1.7149,
      "step": 259
    },
    {
      "epoch": 0.1,
      "learning_rate": 0.00017684011108568592,
      "loss": 1.6936,
      "step": 260
    },
    {
      "epoch": 0.1,
      "learning_rate": 0.00017662805758264893,
      "loss": 1.5013,
      "step": 261
    },
    {
      "epoch": 0.1,
      "learning_rate": 0.00017641516608745114,
      "loss": 1.6229,
      "step": 262
    },
    {
      "epoch": 0.1,
      "learning_rate": 0.00017620143892823977,
      "loss": 1.3457,
      "step": 263
    },
    {
      "epoch": 0.1,
      "learning_rate": 0.00017598687844230088,
      "loss": 1.5029,
      "step": 264
    },
    {
      "epoch": 0.1,
      "learning_rate": 0.0001757714869760335,
      "loss": 1.4726,
      "step": 265
    },
    {
      "epoch": 0.1,
      "learning_rate": 0.0001755552668849242,
      "loss": 1.7338,
      "step": 266
    },
    {
      "epoch": 0.1,
      "learning_rate": 0.00017533822053352128,
      "loss": 1.6824,
      "step": 267
    },
    {
      "epoch": 0.1,
      "learning_rate": 0.00017512035029540885,
      "loss": 1.7167,
      "step": 268
    },
    {
      "epoch": 0.1,
      "learning_rate": 0.00017490165855318094,
      "loss": 1.6222,
      "step": 269
    },
    {
      "epoch": 0.1,
      "learning_rate": 0.0001746821476984154,
      "loss": 1.5473,
      "step": 270
    },
    {
      "epoch": 0.1,
      "learning_rate": 0.00017446182013164778,
      "loss": 1.3884,
      "step": 271
    },
    {
      "epoch": 0.1,
      "learning_rate": 0.000174240678262345,
      "loss": 1.7926,
      "step": 272
    },
    {
      "epoch": 0.1,
      "learning_rate": 0.00017401872450887917,
      "loss": 1.5833,
      "step": 273
    },
    {
      "epoch": 0.1,
      "learning_rate": 0.00017379596129850098,
      "loss": 1.7576,
      "step": 274
    },
    {
      "epoch": 0.1,
      "learning_rate": 0.00017357239106731317,
      "loss": 1.7172,
      "step": 275
    },
    {
      "epoch": 0.1,
      "learning_rate": 0.000173348016260244,
      "loss": 1.4629,
      "step": 276
    },
    {
      "epoch": 0.1,
      "learning_rate": 0.00017312283933102038,
      "loss": 1.5903,
      "step": 277
    },
    {
      "epoch": 0.1,
      "learning_rate": 0.00017289686274214118,
      "loss": 1.8602,
      "step": 278
    },
    {
      "epoch": 0.11,
      "learning_rate": 0.0001726700889648501,
      "loss": 1.6213,
      "step": 279
    },
    {
      "epoch": 0.11,
      "learning_rate": 0.00017244252047910892,
      "loss": 1.7249,
      "step": 280
    },
    {
      "epoch": 0.11,
      "learning_rate": 0.00017221415977357007,
      "loss": 1.6502,
      "step": 281
    },
    {
      "epoch": 0.11,
      "learning_rate": 0.00017198500934554966,
      "loss": 1.5524,
      "step": 282
    },
    {
      "epoch": 0.11,
      "learning_rate": 0.0001717550717010001,
      "loss": 1.4555,
      "step": 283
    },
    {
      "epoch": 0.11,
      "learning_rate": 0.00017152434935448256,
      "loss": 1.5179,
      "step": 284
    },
    {
      "epoch": 0.11,
      "learning_rate": 0.00017129284482913972,
      "loss": 1.8455,
      "step": 285
    },
    {
      "epoch": 0.11,
      "learning_rate": 0.00017106056065666793,
      "loss": 1.3196,
      "step": 286
    },
    {
      "epoch": 0.11,
      "learning_rate": 0.00017082749937728973,
      "loss": 1.5119,
      "step": 287
    },
    {
      "epoch": 0.11,
      "learning_rate": 0.0001705936635397259,
      "loss": 1.5426,
      "step": 288
    },
    {
      "epoch": 0.11,
      "learning_rate": 0.0001703590557011677,
      "loss": 1.5863,
      "step": 289
    },
    {
      "epoch": 0.11,
      "learning_rate": 0.00017012367842724887,
      "loss": 1.6083,
      "step": 290
    },
    {
      "epoch": 0.11,
      "learning_rate": 0.00016988753429201755,
      "loss": 1.4486,
      "step": 291
    },
    {
      "epoch": 0.11,
      "learning_rate": 0.00016965062587790823,
      "loss": 1.5676,
      "step": 292
    },
    {
      "epoch": 0.11,
      "learning_rate": 0.0001694129557757133,
      "loss": 1.6153,
      "step": 293
    },
    {
      "epoch": 0.11,
      "learning_rate": 0.00016917452658455495,
      "loss": 1.8382,
      "step": 294
    },
    {
      "epoch": 0.11,
      "learning_rate": 0.0001689353409118566,
      "loss": 1.7203,
      "step": 295
    },
    {
      "epoch": 0.11,
      "learning_rate": 0.00016869540137331445,
      "loss": 1.8497,
      "step": 296
    },
    {
      "epoch": 0.11,
      "learning_rate": 0.00016845471059286887,
      "loss": 1.6448,
      "step": 297
    },
    {
      "epoch": 0.11,
      "learning_rate": 0.00016821327120267567,
      "loss": 1.7336,
      "step": 298
    },
    {
      "epoch": 0.11,
      "learning_rate": 0.00016797108584307732,
      "loss": 1.9354,
      "step": 299
    },
    {
      "epoch": 0.11,
      "learning_rate": 0.00016772815716257412,
      "loss": 1.388,
      "step": 300
    },
    {
      "epoch": 0.11,
      "learning_rate": 0.0001674844878177952,
      "loss": 1.7663,
      "step": 301
    },
    {
      "epoch": 0.11,
      "learning_rate": 0.00016724008047346947,
      "loss": 1.6741,
      "step": 302
    },
    {
      "epoch": 0.11,
      "learning_rate": 0.0001669949378023965,
      "loss": 1.4152,
      "step": 303
    },
    {
      "epoch": 0.11,
      "learning_rate": 0.00016674906248541726,
      "loss": 1.3268,
      "step": 304
    },
    {
      "epoch": 0.12,
      "learning_rate": 0.0001665024572113848,
      "loss": 1.7646,
      "step": 305
    },
    {
      "epoch": 0.12,
      "learning_rate": 0.000166255124677135,
      "loss": 1.4683,
      "step": 306
    },
    {
      "epoch": 0.12,
      "learning_rate": 0.00016600706758745668,
      "loss": 1.7198,
      "step": 307
    },
    {
      "epoch": 0.12,
      "learning_rate": 0.00016575828865506245,
      "loss": 1.5474,
      "step": 308
    },
    {
      "epoch": 0.12,
      "learning_rate": 0.00016550879060055895,
      "loss": 1.5883,
      "step": 309
    },
    {
      "epoch": 0.12,
      "learning_rate": 0.00016525857615241687,
      "loss": 1.7022,
      "step": 310
    },
    {
      "epoch": 0.12,
      "learning_rate": 0.0001650076480469413,
      "loss": 1.6334,
      "step": 311
    },
    {
      "epoch": 0.12,
      "learning_rate": 0.0001647560090282419,
      "loss": 1.6134,
      "step": 312
    },
    {
      "epoch": 0.12,
      "learning_rate": 0.00016450366184820255,
      "loss": 1.5379,
      "step": 313
    },
    {
      "epoch": 0.12,
      "learning_rate": 0.00016425060926645167,
      "loss": 1.644,
      "step": 314
    },
    {
      "epoch": 0.12,
      "learning_rate": 0.00016399685405033167,
      "loss": 1.7672,
      "step": 315
    },
    {
      "epoch": 0.12,
      "learning_rate": 0.000163742398974869,
      "loss": 1.6279,
      "step": 316
    },
    {
      "epoch": 0.12,
      "learning_rate": 0.00016348724682274353,
      "loss": 1.5003,
      "step": 317
    },
    {
      "epoch": 0.12,
      "learning_rate": 0.00016323140038425842,
      "loss": 1.4426,
      "step": 318
    },
    {
      "epoch": 0.12,
      "learning_rate": 0.00016297486245730927,
      "loss": 1.5677,
      "step": 319
    },
    {
      "epoch": 0.12,
      "learning_rate": 0.0001627176358473537,
      "loss": 1.6226,
      "step": 320
    },
    {
      "epoch": 0.12,
      "learning_rate": 0.0001624597233673808,
      "loss": 1.7272,
      "step": 321
    },
    {
      "epoch": 0.12,
      "learning_rate": 0.0001622011278378801,
      "loss": 1.6675,
      "step": 322
    },
    {
      "epoch": 0.12,
      "learning_rate": 0.00016194185208681083,
      "loss": 1.3257,
      "step": 323
    },
    {
      "epoch": 0.12,
      "learning_rate": 0.0001616818989495711,
      "loss": 1.777,
      "step": 324
    },
    {
      "epoch": 0.12,
      "learning_rate": 0.0001614212712689668,
      "loss": 1.5283,
      "step": 325
    },
    {
      "epoch": 0.12,
      "learning_rate": 0.00016115997189518043,
      "loss": 1.5958,
      "step": 326
    },
    {
      "epoch": 0.12,
      "learning_rate": 0.00016089800368574014,
      "loss": 1.5726,
      "step": 327
    },
    {
      "epoch": 0.12,
      "learning_rate": 0.00016063536950548826,
      "loss": 1.4865,
      "step": 328
    },
    {
      "epoch": 0.12,
      "learning_rate": 0.0001603720722265501,
      "loss": 1.4585,
      "step": 329
    },
    {
      "epoch": 0.12,
      "learning_rate": 0.00016010811472830252,
      "loss": 1.501,
      "step": 330
    },
    {
      "epoch": 0.12,
      "learning_rate": 0.00015984349989734247,
      "loss": 1.4846,
      "step": 331
    },
    {
      "epoch": 0.13,
      "learning_rate": 0.0001595782306274553,
      "loss": 1.5386,
      "step": 332
    },
    {
      "epoch": 0.13,
      "learning_rate": 0.00015931230981958326,
      "loss": 1.4934,
      "step": 333
    },
    {
      "epoch": 0.13,
      "learning_rate": 0.0001590457403817937,
      "loss": 1.7925,
      "step": 334
    },
    {
      "epoch": 0.13,
      "learning_rate": 0.00015877852522924732,
      "loss": 1.7928,
      "step": 335
    },
    {
      "epoch": 0.13,
      "learning_rate": 0.00015851066728416618,
      "loss": 1.6508,
      "step": 336
    },
    {
      "epoch": 0.13,
      "learning_rate": 0.00015824216947580183,
      "loss": 1.6663,
      "step": 337
    },
    {
      "epoch": 0.13,
      "learning_rate": 0.00015797303474040332,
      "loss": 1.8191,
      "step": 338
    },
    {
      "epoch": 0.13,
      "learning_rate": 0.000157703266021185,
      "loss": 1.5659,
      "step": 339
    },
    {
      "epoch": 0.13,
      "learning_rate": 0.00015743286626829437,
      "loss": 1.6147,
      "step": 340
    },
    {
      "epoch": 0.13,
      "learning_rate": 0.00015716183843877976,
      "loss": 1.6888,
      "step": 341
    },
    {
      "epoch": 0.13,
      "learning_rate": 0.00015689018549655813,
      "loss": 1.3531,
      "step": 342
    },
    {
      "epoch": 0.13,
      "learning_rate": 0.00015661791041238256,
      "loss": 1.4753,
      "step": 343
    },
    {
      "epoch": 0.13,
      "learning_rate": 0.00015634501616380967,
      "loss": 1.6138,
      "step": 344
    },
    {
      "epoch": 0.13,
      "learning_rate": 0.0001560715057351673,
      "loss": 1.4431,
      "step": 345
    },
    {
      "epoch": 0.13,
      "learning_rate": 0.00015579738211752165,
      "loss": 1.63,
      "step": 346
    },
    {
      "epoch": 0.13,
      "learning_rate": 0.00015552264830864468,
      "loss": 1.9748,
      "step": 347
    },
    {
      "epoch": 0.13,
      "learning_rate": 0.00015524730731298134,
      "loss": 1.5779,
      "step": 348
    },
    {
      "epoch": 0.13,
      "learning_rate": 0.00015497136214161664,
      "loss": 1.8374,
      "step": 349
    },
    {
      "epoch": 0.13,
      "learning_rate": 0.00015469481581224272,
      "loss": 1.4975,
      "step": 350
    },
    {
      "epoch": 0.13,
      "learning_rate": 0.00015441767134912596,
      "loss": 1.7641,
      "step": 351
    },
    {
      "epoch": 0.13,
      "learning_rate": 0.0001541399317830738,
      "loss": 1.4802,
      "step": 352
    },
    {
      "epoch": 0.13,
      "learning_rate": 0.00015386160015140168,
      "loss": 1.4818,
      "step": 353
    },
    {
      "epoch": 0.13,
      "learning_rate": 0.00015358267949789966,
      "loss": 1.6149,
      "step": 354
    },
    {
      "epoch": 0.13,
      "learning_rate": 0.0001533031728727994,
      "loss": 1.6037,
      "step": 355
    },
    {
      "epoch": 0.13,
      "learning_rate": 0.0001530230833327405,
      "loss": 1.555,
      "step": 356
    },
    {
      "epoch": 0.13,
      "learning_rate": 0.00015274241394073733,
      "loss": 1.8756,
      "step": 357
    },
    {
      "epoch": 0.14,
      "learning_rate": 0.00015246116776614538,
      "loss": 1.7481,
      "step": 358
    },
    {
      "epoch": 0.14,
      "learning_rate": 0.00015217934788462774,
      "loss": 1.7517,
      "step": 359
    },
    {
      "epoch": 0.14,
      "learning_rate": 0.00015189695737812152,
      "loss": 1.6353,
      "step": 360
    },
    {
      "epoch": 0.14,
      "learning_rate": 0.00015161399933480402,
      "loss": 1.5501,
      "step": 361
    },
    {
      "epoch": 0.14,
      "learning_rate": 0.00015133047684905916,
      "loss": 1.5374,
      "step": 362
    },
    {
      "epoch": 0.14,
      "learning_rate": 0.00015104639302144327,
      "loss": 1.5304,
      "step": 363
    },
    {
      "epoch": 0.14,
      "learning_rate": 0.0001507617509586517,
      "loss": 1.4024,
      "step": 364
    },
    {
      "epoch": 0.14,
      "learning_rate": 0.0001504765537734844,
      "loss": 1.6756,
      "step": 365
    },
    {
      "epoch": 0.14,
      "learning_rate": 0.00015019080458481202,
      "loss": 1.7731,
      "step": 366
    },
    {
      "epoch": 0.14,
      "learning_rate": 0.00014990450651754207,
      "loss": 1.7577,
      "step": 367
    },
    {
      "epoch": 0.14,
      "learning_rate": 0.00014961766270258422,
      "loss": 1.6514,
      "step": 368
    },
    {
      "epoch": 0.14,
      "learning_rate": 0.0001493302762768165,
      "loss": 1.5114,
      "step": 369
    },
    {
      "epoch": 0.14,
      "learning_rate": 0.00014904235038305083,
      "loss": 1.5046,
      "step": 370
    },
    {
      "epoch": 0.14,
      "learning_rate": 0.00014875388816999865,
      "loss": 1.6181,
      "step": 371
    },
    {
      "epoch": 0.14,
      "learning_rate": 0.00014846489279223652,
      "loss": 1.622,
      "step": 372
    },
    {
      "epoch": 0.14,
      "learning_rate": 0.00014817536741017152,
      "loss": 1.5497,
      "step": 373
    },
    {
      "epoch": 0.14,
      "learning_rate": 0.00014788531519000696,
      "loss": 1.4466,
      "step": 374
    },
    {
      "epoch": 0.14,
      "learning_rate": 0.00014759473930370736,
      "loss": 1.6712,
      "step": 375
    },
    {
      "epoch": 0.14,
      "learning_rate": 0.0001473036429289641,
      "loss": 1.6282,
      "step": 376
    },
    {
      "epoch": 0.14,
      "learning_rate": 0.0001470120292491605,
      "loss": 1.7498,
      "step": 377
    },
    {
      "epoch": 0.14,
      "learning_rate": 0.00014671990145333696,
      "loss": 1.7531,
      "step": 378
    },
    {
      "epoch": 0.14,
      "learning_rate": 0.0001464272627361564,
      "loss": 1.718,
      "step": 379
    },
    {
      "epoch": 0.14,
      "learning_rate": 0.0001461341162978688,
      "loss": 1.4252,
      "step": 380
    },
    {
      "epoch": 0.14,
      "learning_rate": 0.0001458404653442767,
      "loss": 1.6305,
      "step": 381
    },
    {
      "epoch": 0.14,
      "learning_rate": 0.00014554631308669994,
      "loss": 1.6279,
      "step": 382
    },
    {
      "epoch": 0.14,
      "learning_rate": 0.00014525166274194037,
      "loss": 1.6559,
      "step": 383
    },
    {
      "epoch": 0.14,
      "learning_rate": 0.00014495651753224705,
      "loss": 1.6125,
      "step": 384
    },
    {
      "epoch": 0.15,
      "learning_rate": 0.00014466088068528068,
      "loss": 1.3162,
      "step": 385
    },
    {
      "epoch": 0.15,
      "learning_rate": 0.00014436475543407843,
      "loss": 1.4924,
      "step": 386
    },
    {
      "epoch": 0.15,
      "learning_rate": 0.00014406814501701857,
      "loss": 1.7356,
      "step": 387
    },
    {
      "epoch": 0.15,
      "learning_rate": 0.00014377105267778518,
      "loss": 1.7235,
      "step": 388
    },
    {
      "epoch": 0.15,
      "learning_rate": 0.00014347348166533248,
      "loss": 1.6112,
      "step": 389
    },
    {
      "epoch": 0.15,
      "learning_rate": 0.00014317543523384928,
      "loss": 1.7354,
      "step": 390
    },
    {
      "epoch": 0.15,
      "learning_rate": 0.00014287691664272375,
      "loss": 1.7269,
      "step": 391
    },
    {
      "epoch": 0.15,
      "learning_rate": 0.00014257792915650728,
      "loss": 1.5006,
      "step": 392
    },
    {
      "epoch": 0.15,
      "learning_rate": 0.00014227847604487913,
      "loss": 1.5781,
      "step": 393
    },
    {
      "epoch": 0.15,
      "learning_rate": 0.0001419785605826106,
      "loss": 1.4548,
      "step": 394
    },
    {
      "epoch": 0.15,
      "learning_rate": 0.00014167818604952906,
      "loss": 1.63,
      "step": 395
    },
    {
      "epoch": 0.15,
      "learning_rate": 0.00014137735573048233,
      "loss": 1.5867,
      "step": 396
    },
    {
      "epoch": 0.15,
      "learning_rate": 0.00014107607291530256,
      "loss": 1.5673,
      "step": 397
    },
    {
      "epoch": 0.15,
      "learning_rate": 0.00014077434089877037,
      "loss": 1.6088,
      "step": 398
    },
    {
      "epoch": 0.15,
      "learning_rate": 0.00014047216298057873,
      "loss": 1.7011,
      "step": 399
    },
    {
      "epoch": 0.15,
      "learning_rate": 0.00014016954246529696,
      "loss": 1.6084,
      "step": 400
    },
    {
      "epoch": 0.15,
      "learning_rate": 0.00013986648266233452,
      "loss": 1.7252,
      "step": 401
    },
    {
      "epoch": 0.15,
      "learning_rate": 0.00013956298688590484,
      "loss": 1.6078,
      "step": 402
    },
    {
      "epoch": 0.15,
      "learning_rate": 0.00013925905845498914,
      "loss": 1.7508,
      "step": 403
    },
    {
      "epoch": 0.15,
      "learning_rate": 0.00013895470069330004,
      "loss": 1.3831,
      "step": 404
    },
    {
      "epoch": 0.15,
      "learning_rate": 0.00013864991692924523,
      "loss": 1.5327,
      "step": 405
    },
    {
      "epoch": 0.15,
      "learning_rate": 0.00013834471049589117,
      "loss": 1.4633,
      "step": 406
    },
    {
      "epoch": 0.15,
      "learning_rate": 0.00013803908473092647,
      "loss": 1.6618,
      "step": 407
    },
    {
      "epoch": 0.15,
      "learning_rate": 0.00013773304297662559,
      "loss": 1.7644,
      "step": 408
    },
    {
      "epoch": 0.15,
      "learning_rate": 0.00013742658857981204,
      "loss": 1.4573,
      "step": 409
    },
    {
      "epoch": 0.15,
      "learning_rate": 0.00013711972489182208,
      "loss": 1.5862,
      "step": 410
    },
    {
      "epoch": 0.16,
      "learning_rate": 0.00013681245526846783,
      "loss": 1.5642,
      "step": 411
    },
    {
      "epoch": 0.16,
      "learning_rate": 0.00013650478307000057,
      "loss": 1.6751,
      "step": 412
    },
    {
      "epoch": 0.16,
      "learning_rate": 0.0001361967116610743,
      "loss": 1.6486,
      "step": 413
    },
    {
      "epoch": 0.16,
      "learning_rate": 0.00013588824441070852,
      "loss": 1.6286,
      "step": 414
    },
    {
      "epoch": 0.16,
      "learning_rate": 0.00013557938469225167,
      "loss": 1.5833,
      "step": 415
    },
    {
      "epoch": 0.16,
      "learning_rate": 0.00013527013588334415,
      "loss": 1.6383,
      "step": 416
    },
    {
      "epoch": 0.16,
      "learning_rate": 0.00013496050136588134,
      "loss": 1.7988,
      "step": 417
    },
    {
      "epoch": 0.16,
      "learning_rate": 0.00013465048452597682,
      "loss": 1.5313,
      "step": 418
    },
    {
      "epoch": 0.16,
      "learning_rate": 0.000134340088753925,
      "loss": 1.7617,
      "step": 419
    },
    {
      "epoch": 0.16,
      "learning_rate": 0.00013402931744416433,
      "loss": 1.6456,
      "step": 420
    },
    {
      "epoch": 0.16,
      "learning_rate": 0.00013371817399524005,
      "loss": 1.548,
      "step": 421
    },
    {
      "epoch": 0.16,
      "learning_rate": 0.00013340666180976712,
      "loss": 1.6117,
      "step": 422
    },
    {
      "epoch": 0.16,
      "learning_rate": 0.00013309478429439283,
      "loss": 1.6574,
      "step": 423
    },
    {
      "epoch": 0.16,
      "learning_rate": 0.00013278254485975976,
      "loss": 1.5813,
      "step": 424
    },
    {
      "epoch": 0.16,
      "learning_rate": 0.00013246994692046836,
      "loss": 1.5934,
      "step": 425
    },
    {
      "epoch": 0.16,
      "learning_rate": 0.00013215699389503954,
      "loss": 1.624,
      "step": 426
    },
    {
      "epoch": 0.16,
      "learning_rate": 0.00013184368920587754,
      "loss": 1.4844,
      "step": 427
    },
    {
      "epoch": 0.16,
      "learning_rate": 0.00013153003627923218,
      "loss": 1.5961,
      "step": 428
    },
    {
      "epoch": 0.16,
      "learning_rate": 0.0001312160385451616,
      "loss": 1.7463,
      "step": 429
    },
    {
      "epoch": 0.16,
      "learning_rate": 0.00013090169943749476,
      "loss": 1.6307,
      "step": 430
    },
    {
      "epoch": 0.16,
      "learning_rate": 0.00013058702239379376,
      "loss": 1.7261,
      "step": 431
    },
    {
      "epoch": 0.16,
      "learning_rate": 0.00013027201085531634,
      "loss": 1.733,
      "step": 432
    },
    {
      "epoch": 0.16,
      "learning_rate": 0.00012995666826697819,
      "loss": 1.5627,
      "step": 433
    },
    {
      "epoch": 0.16,
      "learning_rate": 0.0001296409980773154,
      "loss": 1.4446,
      "step": 434
    },
    {
      "epoch": 0.16,
      "learning_rate": 0.0001293250037384465,
      "loss": 1.4851,
      "step": 435
    },
    {
      "epoch": 0.16,
      "learning_rate": 0.00012900868870603503,
      "loss": 1.8002,
      "step": 436
    },
    {
      "epoch": 0.16,
      "learning_rate": 0.0001286920564392514,
      "loss": 1.6335,
      "step": 437
    },
    {
      "epoch": 0.17,
      "learning_rate": 0.0001283751104007355,
      "loss": 1.8301,
      "step": 438
    },
    {
      "epoch": 0.17,
      "learning_rate": 0.00012805785405655833,
      "loss": 1.6114,
      "step": 439
    },
    {
      "epoch": 0.17,
      "learning_rate": 0.00012774029087618446,
      "loss": 1.8021,
      "step": 440
    },
    {
      "epoch": 0.17,
      "learning_rate": 0.00012742242433243396,
      "loss": 1.4632,
      "step": 441
    },
    {
      "epoch": 0.17,
      "learning_rate": 0.00012710425790144446,
      "loss": 1.5362,
      "step": 442
    },
    {
      "epoch": 0.17,
      "learning_rate": 0.00012678579506263297,
      "loss": 1.7287,
      "step": 443
    },
    {
      "epoch": 0.17,
      "learning_rate": 0.00012646703929865817,
      "loss": 1.8786,
      "step": 444
    },
    {
      "epoch": 0.17,
      "learning_rate": 0.00012614799409538198,
      "loss": 1.6777,
      "step": 445
    },
    {
      "epoch": 0.17,
      "learning_rate": 0.00012582866294183167,
      "loss": 1.6491,
      "step": 446
    },
    {
      "epoch": 0.17,
      "learning_rate": 0.00012550904933016153,
      "loss": 1.6196,
      "step": 447
    },
    {
      "epoch": 0.17,
      "learning_rate": 0.00012518915675561483,
      "loss": 1.5774,
      "step": 448
    },
    {
      "epoch": 0.17,
      "learning_rate": 0.0001248689887164855,
      "loss": 1.6325,
      "step": 449
    },
    {
      "epoch": 0.17,
      "learning_rate": 0.00012454854871407994,
      "loss": 1.3833,
      "step": 450
    },
    {
      "epoch": 0.17,
      "learning_rate": 0.00012422784025267864,
      "loss": 1.5203,
      "step": 451
    },
    {
      "epoch": 0.17,
      "learning_rate": 0.00012390686683949798,
      "loss": 1.5659,
      "step": 452
    },
    {
      "epoch": 0.17,
      "learning_rate": 0.00012358563198465182,
      "loss": 1.3192,
      "step": 453
    },
    {
      "epoch": 0.17,
      "learning_rate": 0.00012326413920111303,
      "loss": 1.5701,
      "step": 454
    },
    {
      "epoch": 0.17,
      "learning_rate": 0.00012294239200467516,
      "loss": 1.4334,
      "step": 455
    },
    {
      "epoch": 0.17,
      "learning_rate": 0.00012262039391391404,
      "loss": 1.729,
      "step": 456
    },
    {
      "epoch": 0.17,
      "learning_rate": 0.0001222981484501492,
      "loss": 1.6675,
      "step": 457
    },
    {
      "epoch": 0.17,
      "learning_rate": 0.00012197565913740531,
      "loss": 1.48,
      "step": 458
    },
    {
      "epoch": 0.17,
      "learning_rate": 0.00012165292950237399,
      "loss": 1.533,
      "step": 459
    },
    {
      "epoch": 0.17,
      "learning_rate": 0.0001213299630743747,
      "loss": 1.6758,
      "step": 460
    },
    {
      "epoch": 0.17,
      "learning_rate": 0.0001210067633853166,
      "loss": 1.5528,
      "step": 461
    },
    {
      "epoch": 0.17,
      "learning_rate": 0.00012068333396965968,
      "loss": 1.6283,
      "step": 462
    },
    {
      "epoch": 0.17,
      "learning_rate": 0.00012035967836437625,
      "loss": 1.812,
      "step": 463
    },
    {
      "epoch": 0.18,
      "learning_rate": 0.00012003580010891213,
      "loss": 1.6004,
      "step": 464
    },
    {
      "epoch": 0.18,
      "learning_rate": 0.00011971170274514802,
      "loss": 1.4833,
      "step": 465
    },
    {
      "epoch": 0.18,
      "learning_rate": 0.00011938738981736085,
      "loss": 1.5398,
      "step": 466
    },
    {
      "epoch": 0.18,
      "learning_rate": 0.0001190628648721847,
      "loss": 1.7085,
      "step": 467
    },
    {
      "epoch": 0.18,
      "learning_rate": 0.00011873813145857249,
      "loss": 1.5997,
      "step": 468
    },
    {
      "epoch": 0.18,
      "learning_rate": 0.00011841319312775671,
      "loss": 1.6156,
      "step": 469
    },
    {
      "epoch": 0.18,
      "learning_rate": 0.000118088053433211,
      "loss": 1.4217,
      "step": 470
    },
    {
      "epoch": 0.18,
      "learning_rate": 0.00011776271593061089,
      "loss": 1.5492,
      "step": 471
    },
    {
      "epoch": 0.18,
      "learning_rate": 0.00011743718417779517,
      "loss": 1.6231,
      "step": 472
    },
    {
      "epoch": 0.18,
      "learning_rate": 0.000117111461734727,
      "loss": 1.5459,
      "step": 473
    },
    {
      "epoch": 0.18,
      "learning_rate": 0.00011678555216345477,
      "loss": 1.6095,
      "step": 474
    },
    {
      "epoch": 0.18,
      "learning_rate": 0.00011645945902807341,
      "loss": 1.7221,
      "step": 475
    },
    {
      "epoch": 0.18,
      "learning_rate": 0.00011613318589468511,
      "loss": 1.7852,
      "step": 476
    },
    {
      "epoch": 0.18,
      "learning_rate": 0.00011580673633136065,
      "loss": 1.6094,
      "step": 477
    },
    {
      "epoch": 0.18,
      "learning_rate": 0.00011548011390810017,
      "loss": 1.6245,
      "step": 478
    },
    {
      "epoch": 0.18,
      "learning_rate": 0.00011515332219679404,
      "loss": 1.5804,
      "step": 479
    },
    {
      "epoch": 0.18,
      "learning_rate": 0.0001148263647711842,
      "loss": 1.6381,
      "step": 480
    },
    {
      "epoch": 0.18,
      "learning_rate": 0.0001144992452068246,
      "loss": 1.5489,
      "step": 481
    },
    {
      "epoch": 0.18,
      "learning_rate": 0.00011417196708104243,
      "loss": 1.5808,
      "step": 482
    },
    {
      "epoch": 0.18,
      "learning_rate": 0.00011384453397289876,
      "loss": 1.7598,
      "step": 483
    },
    {
      "epoch": 0.18,
      "learning_rate": 0.0001135169494631497,
      "loss": 1.631,
      "step": 484
    },
    {
      "epoch": 0.18,
      "learning_rate": 0.00011318921713420691,
      "loss": 1.4051,
      "step": 485
    },
    {
      "epoch": 0.18,
      "learning_rate": 0.00011286134057009863,
      "loss": 1.558,
      "step": 486
    },
    {
      "epoch": 0.18,
      "learning_rate": 0.00011253332335643043,
      "loss": 1.7557,
      "step": 487
    },
    {
      "epoch": 0.18,
      "learning_rate": 0.00011220516908034601,
      "loss": 1.3264,
      "step": 488
    },
    {
      "epoch": 0.18,
      "learning_rate": 0.00011187688133048801,
      "loss": 1.513,
      "step": 489
    },
    {
      "epoch": 0.18,
      "learning_rate": 0.00011154846369695863,
      "loss": 1.624,
      "step": 490
    },
    {
      "epoch": 0.19,
      "learning_rate": 0.00011121991977128045,
      "loss": 1.9427,
      "step": 491
    },
    {
      "epoch": 0.19,
      "learning_rate": 0.00011089125314635726,
      "loss": 1.4963,
      "step": 492
    },
    {
      "epoch": 0.19,
      "learning_rate": 0.0001105624674164346,
      "loss": 1.4605,
      "step": 493
    },
    {
      "epoch": 0.19,
      "learning_rate": 0.00011023356617706052,
      "loss": 1.5062,
      "step": 494
    },
    {
      "epoch": 0.19,
      "learning_rate": 0.0001099045530250463,
      "loss": 1.5084,
      "step": 495
    },
    {
      "epoch": 0.19,
      "learning_rate": 0.00010957543155842702,
      "loss": 1.7999,
      "step": 496
    },
    {
      "epoch": 0.19,
      "learning_rate": 0.00010924620537642236,
      "loss": 1.7851,
      "step": 497
    },
    {
      "epoch": 0.19,
      "learning_rate": 0.00010891687807939707,
      "loss": 1.7145,
      "step": 498
    },
    {
      "epoch": 0.19,
      "learning_rate": 0.00010858745326882171,
      "loss": 1.6778,
      "step": 499
    },
    {
      "epoch": 0.19,
      "learning_rate": 0.00010825793454723325,
      "loss": 1.6589,
      "step": 500
    },
    {
      "epoch": 0.19,
      "learning_rate": 0.00010792832551819558,
      "loss": 1.7932,
      "step": 501
    },
    {
      "epoch": 0.19,
      "learning_rate": 0.00010759862978626031,
      "loss": 1.4803,
      "step": 502
    },
    {
      "epoch": 0.19,
      "learning_rate": 0.00010726885095692712,
      "loss": 1.3864,
      "step": 503
    },
    {
      "epoch": 0.19,
      "learning_rate": 0.00010693899263660441,
      "loss": 1.6757,
      "step": 504
    },
    {
      "epoch": 0.19,
      "learning_rate": 0.00010660905843256994,
      "loss": 1.4477,
      "step": 505
    },
    {
      "epoch": 0.19,
      "learning_rate": 0.00010627905195293135,
      "loss": 1.5601,
      "step": 506
    },
    {
      "epoch": 0.19,
      "learning_rate": 0.00010594897680658658,
      "loss": 1.4949,
      "step": 507
    },
    {
      "epoch": 0.19,
      "learning_rate": 0.00010561883660318455,
      "loss": 1.3819,
      "step": 508
    },
    {
      "epoch": 0.19,
      "learning_rate": 0.00010528863495308567,
      "loss": 1.7058,
      "step": 509
    },
    {
      "epoch": 0.19,
      "learning_rate": 0.00010495837546732224,
      "loss": 1.8273,
      "step": 510
    },
    {
      "epoch": 0.19,
      "learning_rate": 0.0001046280617575591,
      "loss": 1.5141,
      "step": 511
    },
    {
      "epoch": 0.19,
      "learning_rate": 0.00010429769743605407,
      "loss": 1.529,
      "step": 512
    },
    {
      "epoch": 0.19,
      "learning_rate": 0.00010396728611561844,
      "loss": 1.6281,
      "step": 513
    },
    {
      "epoch": 0.19,
      "learning_rate": 0.00010363683140957745,
      "loss": 1.5206,
      "step": 514
    },
    {
      "epoch": 0.19,
      "learning_rate": 0.00010330633693173082,
      "loss": 1.7675,
      "step": 515
    },
    {
      "epoch": 0.19,
      "learning_rate": 0.00010297580629631325,
      "loss": 1.6477,
      "step": 516
    },
    {
      "epoch": 0.2,
      "learning_rate": 0.00010264524311795478,
      "loss": 1.5997,
      "step": 517
    },
    {
      "epoch": 0.2,
      "learning_rate": 0.00010231465101164139,
      "loss": 1.8573,
      "step": 518
    },
    {
      "epoch": 0.2,
      "learning_rate": 0.00010198403359267537,
      "loss": 1.4737,
      "step": 519
    },
    {
      "epoch": 0.2,
      "learning_rate": 0.00010165339447663587,
      "loss": 1.7002,
      "step": 520
    },
    {
      "epoch": 0.2,
      "learning_rate": 0.00010132273727933925,
      "loss": 1.5743,
      "step": 521
    },
    {
      "epoch": 0.2,
      "learning_rate": 0.00010099206561679963,
      "loss": 1.6136,
      "step": 522
    },
    {
      "epoch": 0.2,
      "learning_rate": 0.00010066138310518942,
      "loss": 1.5749,
      "step": 523
    },
    {
      "epoch": 0.2,
      "learning_rate": 0.00010033069336079952,
      "loss": 1.7259,
      "step": 524
    },
    {
      "epoch": 0.2,
      "learning_rate": 0.0001,
      "loss": 1.6527,
      "step": 525
    },
    {
      "epoch": 0.2,
      "learning_rate": 9.966930663920049e-05,
      "loss": 1.775,
      "step": 526
    },
    {
      "epoch": 0.2,
      "learning_rate": 9.93386168948106e-05,
      "loss": 1.5532,
      "step": 527
    },
    {
      "epoch": 0.2,
      "learning_rate": 9.900793438320037e-05,
      "loss": 1.7757,
      "step": 528
    },
    {
      "epoch": 0.2,
      "learning_rate": 9.86772627206608e-05,
      "loss": 1.6747,
      "step": 529
    },
    {
      "epoch": 0.2,
      "learning_rate": 9.834660552336415e-05,
      "loss": 1.6145,
      "step": 530
    },
    {
      "epoch": 0.2,
      "learning_rate": 9.801596640732465e-05,
      "loss": 1.7237,
      "step": 531
    },
    {
      "epoch": 0.2,
      "learning_rate": 9.768534898835862e-05,
      "loss": 1.8218,
      "step": 532
    },
    {
      "epoch": 0.2,
      "learning_rate": 9.735475688204521e-05,
      "loss": 1.7219,
      "step": 533
    },
    {
      "epoch": 0.2,
      "learning_rate": 9.702419370368676e-05,
      "loss": 1.6511,
      "step": 534
    },
    {
      "epoch": 0.2,
      "learning_rate": 9.669366306826919e-05,
      "loss": 1.4605,
      "step": 535
    },
    {
      "epoch": 0.2,
      "learning_rate": 9.636316859042259e-05,
      "loss": 1.8221,
      "step": 536
    },
    {
      "epoch": 0.2,
      "learning_rate": 9.603271388438159e-05,
      "loss": 1.4343,
      "step": 537
    },
    {
      "epoch": 0.2,
      "learning_rate": 9.570230256394596e-05,
      "loss": 1.7632,
      "step": 538
    },
    {
      "epoch": 0.2,
      "learning_rate": 9.537193824244091e-05,
      "loss": 1.6369,
      "step": 539
    },
    {
      "epoch": 0.2,
      "learning_rate": 9.504162453267777e-05,
      "loss": 1.8309,
      "step": 540
    },
    {
      "epoch": 0.2,
      "learning_rate": 9.471136504691436e-05,
      "loss": 1.4987,
      "step": 541
    },
    {
      "epoch": 0.2,
      "learning_rate": 9.438116339681545e-05,
      "loss": 1.4305,
      "step": 542
    },
    {
      "epoch": 0.2,
      "learning_rate": 9.405102319341344e-05,
      "loss": 1.6346,
      "step": 543
    },
    {
      "epoch": 0.21,
      "learning_rate": 9.372094804706867e-05,
      "loss": 1.5976,
      "step": 544
    },
    {
      "epoch": 0.21,
      "learning_rate": 9.339094156743007e-05,
      "loss": 1.8561,
      "step": 545
    },
    {
      "epoch": 0.21,
      "learning_rate": 9.30610073633956e-05,
      "loss": 1.6587,
      "step": 546
    },
    {
      "epoch": 0.21,
      "learning_rate": 9.273114904307289e-05,
      "loss": 1.6006,
      "step": 547
    },
    {
      "epoch": 0.21,
      "learning_rate": 9.24013702137397e-05,
      "loss": 1.681,
      "step": 548
    },
    {
      "epoch": 0.21,
      "learning_rate": 9.20716744818044e-05,
      "loss": 1.6353,
      "step": 549
    },
    {
      "epoch": 0.21,
      "learning_rate": 9.174206545276677e-05,
      "loss": 1.5338,
      "step": 550
    },
    {
      "epoch": 0.21,
      "learning_rate": 9.14125467311783e-05,
      "loss": 1.7111,
      "step": 551
    },
    {
      "epoch": 0.21,
      "learning_rate": 9.108312192060298e-05,
      "loss": 1.6232,
      "step": 552
    },
    {
      "epoch": 0.21,
      "learning_rate": 9.075379462357766e-05,
      "loss": 1.5872,
      "step": 553
    },
    {
      "epoch": 0.21,
      "learning_rate": 9.042456844157299e-05,
      "loss": 1.6202,
      "step": 554
    },
    {
      "epoch": 0.21,
      "learning_rate": 9.009544697495374e-05,
      "loss": 1.479,
      "step": 555
    },
    {
      "epoch": 0.21,
      "learning_rate": 8.97664338229395e-05,
      "loss": 1.5796,
      "step": 556
    },
    {
      "epoch": 0.21,
      "learning_rate": 8.943753258356546e-05,
      "loss": 1.7654,
      "step": 557
    },
    {
      "epoch": 0.21,
      "learning_rate": 8.910874685364275e-05,
      "loss": 1.5687,
      "step": 558
    },
    {
      "epoch": 0.21,
      "learning_rate": 8.878008022871958e-05,
      "loss": 1.7592,
      "step": 559
    },
    {
      "epoch": 0.21,
      "learning_rate": 8.845153630304139e-05,
      "loss": 1.6815,
      "step": 560
    },
    {
      "epoch": 0.21,
      "learning_rate": 8.812311866951198e-05,
      "loss": 1.6589,
      "step": 561
    },
    {
      "epoch": 0.21,
      "learning_rate": 8.7794830919654e-05,
      "loss": 1.565,
      "step": 562
    },
    {
      "epoch": 0.21,
      "learning_rate": 8.746667664356956e-05,
      "loss": 1.6184,
      "step": 563
    },
    {
      "epoch": 0.21,
      "learning_rate": 8.713865942990141e-05,
      "loss": 1.6804,
      "step": 564
    },
    {
      "epoch": 0.21,
      "learning_rate": 8.681078286579311e-05,
      "loss": 1.5697,
      "step": 565
    },
    {
      "epoch": 0.21,
      "learning_rate": 8.648305053685034e-05,
      "loss": 1.7056,
      "step": 566
    },
    {
      "epoch": 0.21,
      "learning_rate": 8.615546602710125e-05,
      "loss": 1.6924,
      "step": 567
    },
    {
      "epoch": 0.21,
      "learning_rate": 8.582803291895758e-05,
      "loss": 1.5869,
      "step": 568
    },
    {
      "epoch": 0.21,
      "learning_rate": 8.550075479317542e-05,
      "loss": 1.7831,
      "step": 569
    },
    {
      "epoch": 0.22,
      "learning_rate": 8.517363522881579e-05,
      "loss": 1.7121,
      "step": 570
    },
    {
      "epoch": 0.22,
      "learning_rate": 8.484667780320597e-05,
      "loss": 1.6502,
      "step": 571
    },
    {
      "epoch": 0.22,
      "learning_rate": 8.451988609189987e-05,
      "loss": 1.5674,
      "step": 572
    },
    {
      "epoch": 0.22,
      "learning_rate": 8.419326366863938e-05,
      "loss": 1.7858,
      "step": 573
    },
    {
      "epoch": 0.22,
      "learning_rate": 8.386681410531491e-05,
      "loss": 1.6807,
      "step": 574
    },
    {
      "epoch": 0.22,
      "learning_rate": 8.35405409719266e-05,
      "loss": 1.6116,
      "step": 575
    },
    {
      "epoch": 0.22,
      "learning_rate": 8.321444783654524e-05,
      "loss": 1.6685,
      "step": 576
    },
    {
      "epoch": 0.22,
      "learning_rate": 8.2888538265273e-05,
      "loss": 1.5393,
      "step": 577
    },
    {
      "epoch": 0.22,
      "learning_rate": 8.256281582220485e-05,
      "loss": 1.6817,
      "step": 578
    },
    {
      "epoch": 0.22,
      "learning_rate": 8.223728406938914e-05,
      "loss": 1.8488,
      "step": 579
    },
    {
      "epoch": 0.22,
      "learning_rate": 8.191194656678904e-05,
      "loss": 1.5492,
      "step": 580
    },
    {
      "epoch": 0.22,
      "learning_rate": 8.15868068722433e-05,
      "loss": 1.5958,
      "step": 581
    },
    {
      "epoch": 0.22,
      "learning_rate": 8.126186854142752e-05,
      "loss": 1.6587,
      "step": 582
    },
    {
      "epoch": 0.22,
      "learning_rate": 8.093713512781534e-05,
      "loss": 1.6371,
      "step": 583
    },
    {
      "epoch": 0.22,
      "learning_rate": 8.061261018263919e-05,
      "loss": 1.615,
      "step": 584
    },
    {
      "epoch": 0.22,
      "learning_rate": 8.028829725485199e-05,
      "loss": 1.6166,
      "step": 585
    },
    {
      "epoch": 0.22,
      "learning_rate": 7.996419989108789e-05,
      "loss": 1.5911,
      "step": 586
    },
    {
      "epoch": 0.22,
      "learning_rate": 7.964032163562378e-05,
      "loss": 1.5945,
      "step": 587
    },
    {
      "epoch": 0.22,
      "learning_rate": 7.931666603034033e-05,
      "loss": 1.5537,
      "step": 588
    },
    {
      "epoch": 0.22,
      "learning_rate": 7.899323661468343e-05,
      "loss": 1.4572,
      "step": 589
    },
    {
      "epoch": 0.22,
      "learning_rate": 7.867003692562534e-05,
      "loss": 1.6527,
      "step": 590
    },
    {
      "epoch": 0.22,
      "learning_rate": 7.834707049762603e-05,
      "loss": 1.7753,
      "step": 591
    },
    {
      "epoch": 0.22,
      "learning_rate": 7.80243408625947e-05,
      "loss": 1.5923,
      "step": 592
    },
    {
      "epoch": 0.22,
      "learning_rate": 7.770185154985085e-05,
      "loss": 1.6989,
      "step": 593
    },
    {
      "epoch": 0.22,
      "learning_rate": 7.7379606086086e-05,
      "loss": 1.4455,
      "step": 594
    },
    {
      "epoch": 0.22,
      "learning_rate": 7.705760799532485e-05,
      "loss": 1.654,
      "step": 595
    },
    {
      "epoch": 0.22,
      "learning_rate": 7.673586079888698e-05,
      "loss": 1.8404,
      "step": 596
    },
    {
      "epoch": 0.23,
      "learning_rate": 7.641436801534818e-05,
      "loss": 2.0157,
      "step": 597
    },
    {
      "epoch": 0.23,
      "learning_rate": 7.6093133160502e-05,
      "loss": 1.4572,
      "step": 598
    },
    {
      "epoch": 0.23,
      "learning_rate": 7.577215974732137e-05,
      "loss": 1.6459,
      "step": 599
    },
    {
      "epoch": 0.23,
      "learning_rate": 7.54514512859201e-05,
      "loss": 1.6336,
      "step": 600
    },
    {
      "epoch": 0.23,
      "learning_rate": 7.513101128351454e-05,
      "loss": 1.3708,
      "step": 601
    },
    {
      "epoch": 0.23,
      "learning_rate": 7.48108432443852e-05,
      "loss": 1.4441,
      "step": 602
    },
    {
      "epoch": 0.23,
      "learning_rate": 7.449095066983849e-05,
      "loss": 1.7811,
      "step": 603
    },
    {
      "epoch": 0.23,
      "learning_rate": 7.417133705816837e-05,
      "loss": 1.576,
      "step": 604
    },
    {
      "epoch": 0.23,
      "learning_rate": 7.385200590461803e-05,
      "loss": 1.674,
      "step": 605
    },
    {
      "epoch": 0.23,
      "learning_rate": 7.353296070134186e-05,
      "loss": 1.52,
      "step": 606
    },
    {
      "epoch": 0.23,
      "learning_rate": 7.321420493736705e-05,
      "loss": 1.6435,
      "step": 607
    },
    {
      "epoch": 0.23,
      "learning_rate": 7.289574209855559e-05,
      "loss": 1.6924,
      "step": 608
    },
    {
      "epoch": 0.23,
      "learning_rate": 7.257757566756605e-05,
      "loss": 1.714,
      "step": 609
    },
    {
      "epoch": 0.23,
      "learning_rate": 7.225970912381556e-05,
      "loss": 1.6064,
      "step": 610
    },
    {
      "epoch": 0.23,
      "learning_rate": 7.194214594344168e-05,
      "loss": 1.7011,
      "step": 611
    },
    {
      "epoch": 0.23,
      "learning_rate": 7.16248895992645e-05,
      "loss": 1.5956,
      "step": 612
    },
    {
      "epoch": 0.23,
      "learning_rate": 7.130794356074859e-05,
      "loss": 1.3912,
      "step": 613
    },
    {
      "epoch": 0.23,
      "learning_rate": 7.099131129396501e-05,
      "loss": 1.5384,
      "step": 614
    },
    {
      "epoch": 0.23,
      "learning_rate": 7.067499626155354e-05,
      "loss": 1.7346,
      "step": 615
    },
    {
      "epoch": 0.23,
      "learning_rate": 7.035900192268464e-05,
      "loss": 1.4425,
      "step": 616
    },
    {
      "epoch": 0.23,
      "learning_rate": 7.004333173302185e-05,
      "loss": 1.802,
      "step": 617
    },
    {
      "epoch": 0.23,
      "learning_rate": 6.972798914468369e-05,
      "loss": 1.5367,
      "step": 618
    },
    {
      "epoch": 0.23,
      "learning_rate": 6.941297760620627e-05,
      "loss": 1.5427,
      "step": 619
    },
    {
      "epoch": 0.23,
      "learning_rate": 6.909830056250527e-05,
      "loss": 1.5057,
      "step": 620
    },
    {
      "epoch": 0.23,
      "learning_rate": 6.878396145483841e-05,
      "loss": 1.5293,
      "step": 621
    },
    {
      "epoch": 0.23,
      "learning_rate": 6.846996372076786e-05,
      "loss": 1.4316,
      "step": 622
    },
    {
      "epoch": 0.24,
      "learning_rate": 6.815631079412248e-05,
      "loss": 1.6498,
      "step": 623
    },
    {
      "epoch": 0.24,
      "learning_rate": 6.784300610496048e-05,
      "loss": 1.6795,
      "step": 624
    },
    {
      "epoch": 0.24,
      "learning_rate": 6.753005307953167e-05,
      "loss": 1.4548,
      "step": 625
    },
    {
      "epoch": 0.24,
      "learning_rate": 6.721745514024022e-05,
      "loss": 1.5507,
      "step": 626
    },
    {
      "epoch": 0.24,
      "learning_rate": 6.690521570560717e-05,
      "loss": 1.4182,
      "step": 627
    },
    {
      "epoch": 0.24,
      "learning_rate": 6.65933381902329e-05,
      "loss": 1.7437,
      "step": 628
    },
    {
      "epoch": 0.24,
      "learning_rate": 6.628182600475999e-05,
      "loss": 1.5987,
      "step": 629
    },
    {
      "epoch": 0.24,
      "learning_rate": 6.59706825558357e-05,
      "loss": 1.6205,
      "step": 630
    },
    {
      "epoch": 0.24,
      "learning_rate": 6.565991124607507e-05,
      "loss": 1.6679,
      "step": 631
    },
    {
      "epoch": 0.24,
      "learning_rate": 6.534951547402322e-05,
      "loss": 1.8517,
      "step": 632
    },
    {
      "epoch": 0.24,
      "learning_rate": 6.503949863411865e-05,
      "loss": 1.6901,
      "step": 633
    },
    {
      "epoch": 0.24,
      "learning_rate": 6.47298641166559e-05,
      "loss": 1.5482,
      "step": 634
    },
    {
      "epoch": 0.24,
      "learning_rate": 6.442061530774834e-05,
      "loss": 1.6358,
      "step": 635
    },
    {
      "epoch": 0.24,
      "learning_rate": 6.411175558929152e-05,
      "loss": 1.7407,
      "step": 636
    },
    {
      "epoch": 0.24,
      "learning_rate": 6.38032883389257e-05,
      "loss": 1.7207,
      "step": 637
    },
    {
      "epoch": 0.24,
      "learning_rate": 6.349521692999945e-05,
      "loss": 1.7949,
      "step": 638
    },
    {
      "epoch": 0.24,
      "learning_rate": 6.318754473153221e-05,
      "loss": 1.4887,
      "step": 639
    },
    {
      "epoch": 0.24,
      "learning_rate": 6.28802751081779e-05,
      "loss": 1.4875,
      "step": 640
    },
    {
      "epoch": 0.24,
      "learning_rate": 6.257341142018798e-05,
      "loss": 1.6094,
      "step": 641
    },
    {
      "epoch": 0.24,
      "learning_rate": 6.226695702337442e-05,
      "loss": 1.5121,
      "step": 642
    },
    {
      "epoch": 0.24,
      "learning_rate": 6.196091526907355e-05,
      "loss": 1.7423,
      "step": 643
    },
    {
      "epoch": 0.24,
      "learning_rate": 6.165528950410884e-05,
      "loss": 1.6317,
      "step": 644
    },
    {
      "epoch": 0.24,
      "learning_rate": 6.135008307075481e-05,
      "loss": 1.6648,
      "step": 645
    },
    {
      "epoch": 0.24,
      "learning_rate": 6.10452993067e-05,
      "loss": 1.6393,
      "step": 646
    },
    {
      "epoch": 0.24,
      "learning_rate": 6.074094154501087e-05,
      "loss": 1.5263,
      "step": 647
    },
    {
      "epoch": 0.24,
      "learning_rate": 6.0437013114095195e-05,
      "loss": 1.4215,
      "step": 648
    },
    {
      "epoch": 0.24,
      "learning_rate": 6.01335173376655e-05,
      "loss": 1.5247,
      "step": 649
    },
    {
      "epoch": 0.25,
      "learning_rate": 5.983045753470308e-05,
      "loss": 1.6335,
      "step": 650
    },
    {
      "epoch": 0.25,
      "learning_rate": 5.95278370194213e-05,
      "loss": 1.6662,
      "step": 651
    },
    {
      "epoch": 0.25,
      "learning_rate": 5.922565910122967e-05,
      "loss": 1.6644,
      "step": 652
    },
    {
      "epoch": 0.25,
      "learning_rate": 5.8923927084697475e-05,
      "loss": 1.6358,
      "step": 653
    },
    {
      "epoch": 0.25,
      "learning_rate": 5.862264426951768e-05,
      "loss": 1.2212,
      "step": 654
    },
    {
      "epoch": 0.25,
      "learning_rate": 5.832181395047098e-05,
      "loss": 1.5045,
      "step": 655
    },
    {
      "epoch": 0.25,
      "learning_rate": 5.8021439417389444e-05,
      "loss": 1.5293,
      "step": 656
    },
    {
      "epoch": 0.25,
      "learning_rate": 5.772152395512087e-05,
      "loss": 1.6957,
      "step": 657
    },
    {
      "epoch": 0.25,
      "learning_rate": 5.7422070843492734e-05,
      "loss": 1.6455,
      "step": 658
    },
    {
      "epoch": 0.25,
      "learning_rate": 5.7123083357276283e-05,
      "loss": 1.7537,
      "step": 659
    },
    {
      "epoch": 0.25,
      "learning_rate": 5.6824564766150726e-05,
      "loss": 1.369,
      "step": 660
    },
    {
      "epoch": 0.25,
      "learning_rate": 5.652651833466755e-05,
      "loss": 1.5874,
      "step": 661
    },
    {
      "epoch": 0.25,
      "learning_rate": 5.622894732221482e-05,
      "loss": 1.5345,
      "step": 662
    },
    {
      "epoch": 0.25,
      "learning_rate": 5.593185498298141e-05,
      "loss": 1.4931,
      "step": 663
    },
    {
      "epoch": 0.25,
      "learning_rate": 5.563524456592163e-05,
      "loss": 1.8471,
      "step": 664
    },
    {
      "epoch": 0.25,
      "learning_rate": 5.533911931471936e-05,
      "loss": 1.512,
      "step": 665
    },
    {
      "epoch": 0.25,
      "learning_rate": 5.504348246775299e-05,
      "loss": 1.6631,
      "step": 666
    },
    {
      "epoch": 0.25,
      "learning_rate": 5.4748337258059626e-05,
      "loss": 1.5525,
      "step": 667
    },
    {
      "epoch": 0.25,
      "learning_rate": 5.4453686913300074e-05,
      "loss": 1.8276,
      "step": 668
    },
    {
      "epoch": 0.25,
      "learning_rate": 5.415953465572332e-05,
      "loss": 1.6255,
      "step": 669
    },
    {
      "epoch": 0.25,
      "learning_rate": 5.386588370213124e-05,
      "loss": 1.5748,
      "step": 670
    },
    {
      "epoch": 0.25,
      "learning_rate": 5.357273726384367e-05,
      "loss": 1.7415,
      "step": 671
    },
    {
      "epoch": 0.25,
      "learning_rate": 5.328009854666303e-05,
      "loss": 1.2476,
      "step": 672
    },
    {
      "epoch": 0.25,
      "learning_rate": 5.2987970750839555e-05,
      "loss": 1.4693,
      "step": 673
    },
    {
      "epoch": 0.25,
      "learning_rate": 5.269635707103593e-05,
      "loss": 1.5491,
      "step": 674
    },
    {
      "epoch": 0.25,
      "learning_rate": 5.240526069629265e-05,
      "loss": 1.5098,
      "step": 675
    },
    {
      "epoch": 0.26,
      "learning_rate": 5.2114684809993044e-05,
      "loss": 1.6815,
      "step": 676
    },
    {
      "epoch": 0.26,
      "learning_rate": 5.182463258982846e-05,
      "loss": 1.5905,
      "step": 677
    },
    {
      "epoch": 0.26,
      "learning_rate": 5.1535107207763534e-05,
      "loss": 1.5357,
      "step": 678
    },
    {
      "epoch": 0.26,
      "learning_rate": 5.124611183000138e-05,
      "loss": 1.7417,
      "step": 679
    },
    {
      "epoch": 0.26,
      "learning_rate": 5.095764961694922e-05,
      "loss": 1.7145,
      "step": 680
    },
    {
      "epoch": 0.26,
      "learning_rate": 5.0669723723183506e-05,
      "loss": 1.6636,
      "step": 681
    },
    {
      "epoch": 0.26,
      "learning_rate": 5.0382337297415773e-05,
      "loss": 1.4667,
      "step": 682
    },
    {
      "epoch": 0.26,
      "learning_rate": 5.0095493482457955e-05,
      "loss": 1.6005,
      "step": 683
    },
    {
      "epoch": 0.26,
      "learning_rate": 4.980919541518796e-05,
      "loss": 1.7212,
      "step": 684
    },
    {
      "epoch": 0.26,
      "learning_rate": 4.952344622651566e-05,
      "loss": 1.6054,
      "step": 685
    },
    {
      "epoch": 0.26,
      "learning_rate": 4.923824904134829e-05,
      "loss": 1.4423,
      "step": 686
    },
    {
      "epoch": 0.26,
      "learning_rate": 4.895360697855674e-05,
      "loss": 1.5751,
      "step": 687
    },
    {
      "epoch": 0.26,
      "learning_rate": 4.866952315094088e-05,
      "loss": 1.3508,
      "step": 688
    },
    {
      "epoch": 0.26,
      "learning_rate": 4.838600066519596e-05,
      "loss": 1.5055,
      "step": 689
    },
    {
      "epoch": 0.26,
      "learning_rate": 4.810304262187852e-05,
      "loss": 1.5928,
      "step": 690
    },
    {
      "epoch": 0.26,
      "learning_rate": 4.7820652115372253e-05,
      "loss": 1.5872,
      "step": 691
    },
    {
      "epoch": 0.26,
      "learning_rate": 4.753883223385467e-05,
      "loss": 1.6046,
      "step": 692
    },
    {
      "epoch": 0.26,
      "learning_rate": 4.72575860592627e-05,
      "loss": 1.5913,
      "step": 693
    },
    {
      "epoch": 0.26,
      "learning_rate": 4.697691666725955e-05,
      "loss": 1.6866,
      "step": 694
    },
    {
      "epoch": 0.26,
      "learning_rate": 4.669682712720065e-05,
      "loss": 1.5976,
      "step": 695
    },
    {
      "epoch": 0.26,
      "learning_rate": 4.6417320502100316e-05,
      "loss": 1.4559,
      "step": 696
    },
    {
      "epoch": 0.26,
      "learning_rate": 4.613839984859835e-05,
      "loss": 1.6007,
      "step": 697
    },
    {
      "epoch": 0.26,
      "learning_rate": 4.58600682169262e-05,
      "loss": 1.5832,
      "step": 698
    },
    {
      "epoch": 0.26,
      "learning_rate": 4.55823286508741e-05,
      "loss": 1.6939,
      "step": 699
    },
    {
      "epoch": 0.26,
      "learning_rate": 4.530518418775733e-05,
      "loss": 1.5553,
      "step": 700
    },
    {
      "epoch": 0.26,
      "learning_rate": 4.502863785838342e-05,
      "loss": 1.4034,
      "step": 701
    },
    {
      "epoch": 0.26,
      "learning_rate": 4.475269268701868e-05,
      "loss": 1.4394,
      "step": 702
    },
    {
      "epoch": 0.27,
      "learning_rate": 4.447735169135533e-05,
      "loss": 1.6641,
      "step": 703
    },
    {
      "epoch": 0.27,
      "learning_rate": 4.4202617882478405e-05,
      "loss": 1.4679,
      "step": 704
    },
    {
      "epoch": 0.27,
      "learning_rate": 4.392849426483274e-05,
      "loss": 1.5268,
      "step": 705
    },
    {
      "epoch": 0.27,
      "learning_rate": 4.365498383619036e-05,
      "loss": 1.8219,
      "step": 706
    },
    {
      "epoch": 0.27,
      "learning_rate": 4.338208958761747e-05,
      "loss": 1.4109,
      "step": 707
    },
    {
      "epoch": 0.27,
      "learning_rate": 4.310981450344189e-05,
      "loss": 1.4852,
      "step": 708
    },
    {
      "epoch": 0.27,
      "learning_rate": 4.2838161561220245e-05,
      "loss": 1.7155,
      "step": 709
    },
    {
      "epoch": 0.27,
      "learning_rate": 4.256713373170564e-05,
      "loss": 1.6242,
      "step": 710
    },
    {
      "epoch": 0.27,
      "learning_rate": 4.229673397881499e-05,
      "loss": 1.49,
      "step": 711
    },
    {
      "epoch": 0.27,
      "learning_rate": 4.2026965259596666e-05,
      "loss": 1.5365,
      "step": 712
    },
    {
      "epoch": 0.27,
      "learning_rate": 4.1757830524198195e-05,
      "loss": 1.5405,
      "step": 713
    },
    {
      "epoch": 0.27,
      "learning_rate": 4.148933271583385e-05,
      "loss": 1.7893,
      "step": 714
    },
    {
      "epoch": 0.27,
      "learning_rate": 4.12214747707527e-05,
      "loss": 1.6398,
      "step": 715
    },
    {
      "epoch": 0.27,
      "learning_rate": 4.0954259618206295e-05,
      "loss": 1.1292,
      "step": 716
    },
    {
      "epoch": 0.27,
      "learning_rate": 4.0687690180416736e-05,
      "loss": 1.7568,
      "step": 717
    },
    {
      "epoch": 0.27,
      "learning_rate": 4.0421769372544736e-05,
      "loss": 1.6105,
      "step": 718
    },
    {
      "epoch": 0.27,
      "learning_rate": 4.015650010265757e-05,
      "loss": 1.6282,
      "step": 719
    },
    {
      "epoch": 0.27,
      "learning_rate": 3.9891885271697496e-05,
      "loss": 1.4906,
      "step": 720
    },
    {
      "epoch": 0.27,
      "learning_rate": 3.962792777344992e-05,
      "loss": 1.6453,
      "step": 721
    },
    {
      "epoch": 0.27,
      "learning_rate": 3.936463049451179e-05,
      "loss": 1.5742,
      "step": 722
    },
    {
      "epoch": 0.27,
      "learning_rate": 3.910199631425989e-05,
      "loss": 1.7459,
      "step": 723
    },
    {
      "epoch": 0.27,
      "learning_rate": 3.884002810481958e-05,
      "loss": 1.4263,
      "step": 724
    },
    {
      "epoch": 0.27,
      "learning_rate": 3.857872873103322e-05,
      "loss": 1.6211,
      "step": 725
    },
    {
      "epoch": 0.27,
      "learning_rate": 3.8318101050428904e-05,
      "loss": 1.6104,
      "step": 726
    },
    {
      "epoch": 0.27,
      "learning_rate": 3.805814791318921e-05,
      "loss": 1.5656,
      "step": 727
    },
    {
      "epoch": 0.27,
      "learning_rate": 3.779887216211995e-05,
      "loss": 1.5755,
      "step": 728
    },
    {
      "epoch": 0.28,
      "learning_rate": 3.754027663261922e-05,
      "loss": 1.5348,
      "step": 729
    },
    {
      "epoch": 0.28,
      "learning_rate": 3.7282364152646297e-05,
      "loss": 1.5228,
      "step": 730
    },
    {
      "epoch": 0.28,
      "learning_rate": 3.702513754269076e-05,
      "loss": 1.2992,
      "step": 731
    },
    {
      "epoch": 0.28,
      "learning_rate": 3.676859961574162e-05,
      "loss": 1.7796,
      "step": 732
    },
    {
      "epoch": 0.28,
      "learning_rate": 3.6512753177256476e-05,
      "loss": 1.3485,
      "step": 733
    },
    {
      "epoch": 0.28,
      "learning_rate": 3.6257601025131026e-05,
      "loss": 1.4678,
      "step": 734
    },
    {
      "epoch": 0.28,
      "learning_rate": 3.600314594966834e-05,
      "loss": 1.244,
      "step": 735
    },
    {
      "epoch": 0.28,
      "learning_rate": 3.574939073354838e-05,
      "loss": 1.7183,
      "step": 736
    },
    {
      "epoch": 0.28,
      "learning_rate": 3.549633815179746e-05,
      "loss": 1.473,
      "step": 737
    },
    {
      "epoch": 0.28,
      "learning_rate": 3.5243990971758125e-05,
      "loss": 1.6696,
      "step": 738
    },
    {
      "epoch": 0.28,
      "learning_rate": 3.499235195305868e-05,
      "loss": 1.4558,
      "step": 739
    },
    {
      "epoch": 0.28,
      "learning_rate": 3.4741423847583134e-05,
      "loss": 1.5574,
      "step": 740
    },
    {
      "epoch": 0.28,
      "learning_rate": 3.449120939944107e-05,
      "loss": 1.7925,
      "step": 741
    },
    {
      "epoch": 0.28,
      "learning_rate": 3.424171134493756e-05,
      "loss": 1.5674,
      "step": 742
    },
    {
      "epoch": 0.28,
      "learning_rate": 3.399293241254336e-05,
      "loss": 1.689,
      "step": 743
    },
    {
      "epoch": 0.28,
      "learning_rate": 3.3744875322865034e-05,
      "loss": 1.3738,
      "step": 744
    },
    {
      "epoch": 0.28,
      "learning_rate": 3.349754278861517e-05,
      "loss": 1.4457,
      "step": 745
    },
    {
      "epoch": 0.28,
      "learning_rate": 3.325093751458276e-05,
      "loss": 1.7723,
      "step": 746
    },
    {
      "epoch": 0.28,
      "learning_rate": 3.3005062197603506e-05,
      "loss": 1.5807,
      "step": 747
    },
    {
      "epoch": 0.28,
      "learning_rate": 3.275991952653054e-05,
      "loss": 1.69,
      "step": 748
    },
    {
      "epoch": 0.28,
      "learning_rate": 3.25155121822048e-05,
      "loss": 1.5512,
      "step": 749
    },
    {
      "epoch": 0.28,
      "learning_rate": 3.227184283742591e-05,
      "loss": 1.4606,
      "step": 750
    },
    {
      "epoch": 0.28,
      "learning_rate": 3.2028914156922705e-05,
      "loss": 1.6702,
      "step": 751
    },
    {
      "epoch": 0.28,
      "learning_rate": 3.178672879732435e-05,
      "loss": 1.5935,
      "step": 752
    },
    {
      "epoch": 0.28,
      "learning_rate": 3.154528940713113e-05,
      "loss": 1.8427,
      "step": 753
    },
    {
      "epoch": 0.28,
      "learning_rate": 3.1304598626685545e-05,
      "loss": 1.5779,
      "step": 754
    },
    {
      "epoch": 0.28,
      "learning_rate": 3.106465908814342e-05,
      "loss": 1.4777,
      "step": 755
    },
    {
      "epoch": 0.29,
      "learning_rate": 3.0825473415445074e-05,
      "loss": 1.7201,
      "step": 756
    },
    {
      "epoch": 0.29,
      "learning_rate": 3.0587044224286746e-05,
      "loss": 1.557,
      "step": 757
    },
    {
      "epoch": 0.29,
      "learning_rate": 3.034937412209178e-05,
      "loss": 1.4025,
      "step": 758
    },
    {
      "epoch": 0.29,
      "learning_rate": 3.0112465707982417e-05,
      "loss": 1.4136,
      "step": 759
    },
    {
      "epoch": 0.29,
      "learning_rate": 2.9876321572751144e-05,
      "loss": 1.602,
      "step": 760
    },
    {
      "epoch": 0.29,
      "learning_rate": 2.964094429883231e-05,
      "loss": 1.4762,
      "step": 761
    },
    {
      "epoch": 0.29,
      "learning_rate": 2.940633646027414e-05,
      "loss": 1.5751,
      "step": 762
    },
    {
      "epoch": 0.29,
      "learning_rate": 2.9172500622710263e-05,
      "loss": 1.4926,
      "step": 763
    },
    {
      "epoch": 0.29,
      "learning_rate": 2.8939439343332086e-05,
      "loss": 1.5308,
      "step": 764
    },
    {
      "epoch": 0.29,
      "learning_rate": 2.87071551708603e-05,
      "loss": 1.2756,
      "step": 765
    },
    {
      "epoch": 0.29,
      "learning_rate": 2.8475650645517472e-05,
      "loss": 1.5544,
      "step": 766
    },
    {
      "epoch": 0.29,
      "learning_rate": 2.824492829899994e-05,
      "loss": 1.43,
      "step": 767
    },
    {
      "epoch": 0.29,
      "learning_rate": 2.8014990654450325e-05,
      "loss": 1.3823,
      "step": 768
    },
    {
      "epoch": 0.29,
      "learning_rate": 2.778584022642996e-05,
      "loss": 1.5595,
      "step": 769
    },
    {
      "epoch": 0.29,
      "learning_rate": 2.7557479520891104e-05,
      "loss": 1.6086,
      "step": 770
    },
    {
      "epoch": 0.29,
      "learning_rate": 2.7329911035149937e-05,
      "loss": 1.5887,
      "step": 771
    },
    {
      "epoch": 0.29,
      "learning_rate": 2.7103137257858868e-05,
      "loss": 1.5278,
      "step": 772
    },
    {
      "epoch": 0.29,
      "learning_rate": 2.687716066897964e-05,
      "loss": 1.7046,
      "step": 773
    },
    {
      "epoch": 0.29,
      "learning_rate": 2.6651983739756026e-05,
      "loss": 1.5074,
      "step": 774
    },
    {
      "epoch": 0.29,
      "learning_rate": 2.6427608932686843e-05,
      "loss": 1.5464,
      "step": 775
    },
    {
      "epoch": 0.29,
      "learning_rate": 2.6204038701499056e-05,
      "loss": 1.4047,
      "step": 776
    },
    {
      "epoch": 0.29,
      "learning_rate": 2.598127549112084e-05,
      "loss": 1.5001,
      "step": 777
    },
    {
      "epoch": 0.29,
      "learning_rate": 2.5759321737655017e-05,
      "loss": 1.7997,
      "step": 778
    },
    {
      "epoch": 0.29,
      "learning_rate": 2.5538179868352253e-05,
      "loss": 1.8855,
      "step": 779
    },
    {
      "epoch": 0.29,
      "learning_rate": 2.5317852301584643e-05,
      "loss": 1.5479,
      "step": 780
    },
    {
      "epoch": 0.29,
      "learning_rate": 2.5098341446819097e-05,
      "loss": 1.6239,
      "step": 781
    },
    {
      "epoch": 0.3,
      "learning_rate": 2.487964970459118e-05,
      "loss": 1.3434,
      "step": 782
    },
    {
      "epoch": 0.3,
      "learning_rate": 2.466177946647874e-05,
      "loss": 1.4003,
      "step": 783
    },
    {
      "epoch": 0.3,
      "learning_rate": 2.4444733115075823e-05,
      "loss": 1.3932,
      "step": 784
    },
    {
      "epoch": 0.3,
      "learning_rate": 2.422851302396655e-05,
      "loss": 1.4909,
      "step": 785
    },
    {
      "epoch": 0.3,
      "learning_rate": 2.4013121557699157e-05,
      "loss": 1.5588,
      "step": 786
    },
    {
      "epoch": 0.3,
      "learning_rate": 2.3798561071760238e-05,
      "loss": 1.6942,
      "step": 787
    },
    {
      "epoch": 0.3,
      "learning_rate": 2.3584833912548888e-05,
      "loss": 1.4435,
      "step": 788
    },
    {
      "epoch": 0.3,
      "learning_rate": 2.3371942417351077e-05,
      "loss": 1.8429,
      "step": 789
    },
    {
      "epoch": 0.3,
      "learning_rate": 2.315988891431412e-05,
      "loss": 1.5738,
      "step": 790
    },
    {
      "epoch": 0.3,
      "learning_rate": 2.2948675722421086e-05,
      "loss": 1.5925,
      "step": 791
    },
    {
      "epoch": 0.3,
      "learning_rate": 2.2738305151465645e-05,
      "loss": 1.6478,
      "step": 792
    },
    {
      "epoch": 0.3,
      "learning_rate": 2.2528779502026652e-05,
      "loss": 1.5971,
      "step": 793
    },
    {
      "epoch": 0.3,
      "learning_rate": 2.2320101065443056e-05,
      "loss": 1.3759,
      "step": 794
    },
    {
      "epoch": 0.3,
      "learning_rate": 2.2112272123788768e-05,
      "loss": 1.5112,
      "step": 795
    },
    {
      "epoch": 0.3,
      "learning_rate": 2.190529494984782e-05,
      "loss": 1.4325,
      "step": 796
    },
    {
      "epoch": 0.3,
      "learning_rate": 2.1699171807089412e-05,
      "loss": 1.3526,
      "step": 797
    },
    {
      "epoch": 0.3,
      "learning_rate": 2.149390494964323e-05,
      "loss": 1.4915,
      "step": 798
    },
    {
      "epoch": 0.3,
      "learning_rate": 2.1289496622274753e-05,
      "loss": 1.928,
      "step": 799
    },
    {
      "epoch": 0.3,
      "learning_rate": 2.1085949060360654e-05,
      "loss": 1.6015,
      "step": 800
    },
    {
      "epoch": 0.3,
      "learning_rate": 2.0883264489864475e-05,
      "loss": 1.7447,
      "step": 801
    },
    {
      "epoch": 0.3,
      "learning_rate": 2.0681445127312214e-05,
      "loss": 1.4804,
      "step": 802
    },
    {
      "epoch": 0.3,
      "learning_rate": 2.048049317976809e-05,
      "loss": 1.7259,
      "step": 803
    },
    {
      "epoch": 0.3,
      "learning_rate": 2.0280410844810428e-05,
      "loss": 1.507,
      "step": 804
    },
    {
      "epoch": 0.3,
      "learning_rate": 2.008120031050753e-05,
      "loss": 1.5448,
      "step": 805
    },
    {
      "epoch": 0.3,
      "learning_rate": 1.988286375539391e-05,
      "loss": 1.414,
      "step": 806
    },
    {
      "epoch": 0.3,
      "learning_rate": 1.9685403348446374e-05,
      "loss": 1.3942,
      "step": 807
    },
    {
      "epoch": 0.3,
      "learning_rate": 1.9488821249060297e-05,
      "loss": 1.4963,
      "step": 808
    },
    {
      "epoch": 0.31,
      "learning_rate": 1.9293119607025987e-05,
      "loss": 1.657,
      "step": 809
    },
    {
      "epoch": 0.31,
      "learning_rate": 1.9098300562505266e-05,
      "loss": 1.4615,
      "step": 810
    },
    {
      "epoch": 0.31,
      "learning_rate": 1.8904366246008e-05,
      "loss": 1.6224,
      "step": 811
    },
    {
      "epoch": 0.31,
      "learning_rate": 1.871131877836879e-05,
      "loss": 1.7225,
      "step": 812
    },
    {
      "epoch": 0.31,
      "learning_rate": 1.8519160270723857e-05,
      "loss": 1.3401,
      "step": 813
    },
    {
      "epoch": 0.31,
      "learning_rate": 1.8327892824487792e-05,
      "loss": 1.4441,
      "step": 814
    },
    {
      "epoch": 0.31,
      "learning_rate": 1.8137518531330767e-05,
      "loss": 1.7494,
      "step": 815
    },
    {
      "epoch": 0.31,
      "learning_rate": 1.7948039473155554e-05,
      "loss": 1.6001,
      "step": 816
    },
    {
      "epoch": 0.31,
      "learning_rate": 1.7759457722074768e-05,
      "loss": 1.5838,
      "step": 817
    },
    {
      "epoch": 0.31,
      "learning_rate": 1.7571775340388276e-05,
      "loss": 1.5504,
      "step": 818
    },
    {
      "epoch": 0.31,
      "learning_rate": 1.738499438056045e-05,
      "loss": 1.6978,
      "step": 819
    },
    {
      "epoch": 0.31,
      "learning_rate": 1.7199116885197995e-05,
      "loss": 1.6522,
      "step": 820
    },
    {
      "epoch": 0.31,
      "learning_rate": 1.7014144887027404e-05,
      "loss": 1.5064,
      "step": 821
    },
    {
      "epoch": 0.31,
      "learning_rate": 1.683008040887285e-05,
      "loss": 1.8634,
      "step": 822
    },
    {
      "epoch": 0.31,
      "learning_rate": 1.6646925463633922e-05,
      "loss": 1.4848,
      "step": 823
    },
    {
      "epoch": 0.31,
      "learning_rate": 1.646468205426377e-05,
      "loss": 1.7141,
      "step": 824
    },
    {
      "epoch": 0.31,
      "learning_rate": 1.6283352173747145e-05,
      "loss": 1.7473,
      "step": 825
    },
    {
      "epoch": 0.31,
      "learning_rate": 1.6102937805078544e-05,
      "loss": 1.3034,
      "step": 826
    },
    {
      "epoch": 0.31,
      "learning_rate": 1.592344092124064e-05,
      "loss": 1.4523,
      "step": 827
    },
    {
      "epoch": 0.31,
      "learning_rate": 1.5744863485182537e-05,
      "loss": 1.6129,
      "step": 828
    },
    {
      "epoch": 0.31,
      "learning_rate": 1.5567207449798515e-05,
      "loss": 1.4829,
      "step": 829
    },
    {
      "epoch": 0.31,
      "learning_rate": 1.5390474757906446e-05,
      "loss": 1.5858,
      "step": 830
    },
    {
      "epoch": 0.31,
      "learning_rate": 1.5214667342226818e-05,
      "loss": 1.5967,
      "step": 831
    },
    {
      "epoch": 0.31,
      "learning_rate": 1.5039787125361326e-05,
      "loss": 1.6712,
      "step": 832
    },
    {
      "epoch": 0.31,
      "learning_rate": 1.4865836019771995e-05,
      "loss": 1.6599,
      "step": 833
    },
    {
      "epoch": 0.31,
      "learning_rate": 1.4692815927760273e-05,
      "loss": 1.6611,
      "step": 834
    },
    {
      "epoch": 0.32,
      "learning_rate": 1.4520728741446089e-05,
      "loss": 1.5227,
      "step": 835
    },
    {
      "epoch": 0.32,
      "learning_rate": 1.4349576342747462e-05,
      "loss": 1.752,
      "step": 836
    },
    {
      "epoch": 0.32,
      "learning_rate": 1.4179360603359504e-05,
      "loss": 1.8085,
      "step": 837
    },
    {
      "epoch": 0.32,
      "learning_rate": 1.4010083384734308e-05,
      "loss": 1.5195,
      "step": 838
    },
    {
      "epoch": 0.32,
      "learning_rate": 1.384174653806044e-05,
      "loss": 1.4271,
      "step": 839
    },
    {
      "epoch": 0.32,
      "learning_rate": 1.3674351904242611e-05,
      "loss": 1.5761,
      "step": 840
    },
    {
      "epoch": 0.32,
      "learning_rate": 1.350790131388181e-05,
      "loss": 1.2839,
      "step": 841
    },
    {
      "epoch": 0.32,
      "learning_rate": 1.3342396587254958e-05,
      "loss": 1.6232,
      "step": 842
    },
    {
      "epoch": 0.32,
      "learning_rate": 1.3177839534295277e-05,
      "loss": 1.5391,
      "step": 843
    },
    {
      "epoch": 0.32,
      "learning_rate": 1.3014231954572287e-05,
      "loss": 1.5893,
      "step": 844
    },
    {
      "epoch": 0.32,
      "learning_rate": 1.2851575637272262e-05,
      "loss": 1.5226,
      "step": 845
    },
    {
      "epoch": 0.32,
      "learning_rate": 1.2689872361178701e-05,
      "loss": 1.3675,
      "step": 846
    },
    {
      "epoch": 0.32,
      "learning_rate": 1.2529123894652661e-05,
      "loss": 1.4554,
      "step": 847
    },
    {
      "epoch": 0.32,
      "learning_rate": 1.2369331995613665e-05,
      "loss": 1.2827,
      "step": 848
    },
    {
      "epoch": 0.32,
      "learning_rate": 1.2210498411520255e-05,
      "loss": 1.7629,
      "step": 849
    },
    {
      "epoch": 0.32,
      "learning_rate": 1.2052624879351104e-05,
      "loss": 1.3588,
      "step": 850
    },
    {
      "epoch": 0.32,
      "learning_rate": 1.189571312558585e-05,
      "loss": 1.4657,
      "step": 851
    },
    {
      "epoch": 0.32,
      "learning_rate": 1.173976486618631e-05,
      "loss": 1.7873,
      "step": 852
    },
    {
      "epoch": 0.32,
      "learning_rate": 1.158478180657766e-05,
      "loss": 1.6978,
      "step": 853
    },
    {
      "epoch": 0.32,
      "learning_rate": 1.143076564162977e-05,
      "loss": 1.4346,
      "step": 854
    },
    {
      "epoch": 0.32,
      "learning_rate": 1.1277718055638819e-05,
      "loss": 1.4188,
      "step": 855
    },
    {
      "epoch": 0.32,
      "learning_rate": 1.1125640722308628e-05,
      "loss": 1.4854,
      "step": 856
    },
    {
      "epoch": 0.32,
      "learning_rate": 1.097453530473258e-05,
      "loss": 1.5549,
      "step": 857
    },
    {
      "epoch": 0.32,
      "learning_rate": 1.0824403455375288e-05,
      "loss": 1.4872,
      "step": 858
    },
    {
      "epoch": 0.32,
      "learning_rate": 1.0675246816054586e-05,
      "loss": 1.45,
      "step": 859
    },
    {
      "epoch": 0.32,
      "learning_rate": 1.0527067017923654e-05,
      "loss": 1.4492,
      "step": 860
    },
    {
      "epoch": 0.32,
      "learning_rate": 1.0379865681452971e-05,
      "loss": 1.5337,
      "step": 861
    },
    {
      "epoch": 0.33,
      "learning_rate": 1.0233644416412791e-05,
      "loss": 1.6242,
      "step": 862
    },
    {
      "epoch": 0.33,
      "learning_rate": 1.0088404821855412e-05,
      "loss": 1.4762,
      "step": 863
    },
    {
      "epoch": 0.33,
      "learning_rate": 9.944148486097793e-06,
      "loss": 1.3621,
      "step": 864
    },
    {
      "epoch": 0.33,
      "learning_rate": 9.80087698670411e-06,
      "loss": 1.8927,
      "step": 865
    },
    {
      "epoch": 0.33,
      "learning_rate": 9.658591890468515e-06,
      "loss": 1.699,
      "step": 866
    },
    {
      "epoch": 0.33,
      "learning_rate": 9.517294753398064e-06,
      "loss": 1.5539,
      "step": 867
    },
    {
      "epoch": 0.33,
      "learning_rate": 9.376987120695545e-06,
      "loss": 1.4888,
      "step": 868
    },
    {
      "epoch": 0.33,
      "learning_rate": 9.237670526742793e-06,
      "loss": 1.5137,
      "step": 869
    },
    {
      "epoch": 0.33,
      "learning_rate": 9.09934649508375e-06,
      "loss": 1.3459,
      "step": 870
    },
    {
      "epoch": 0.33,
      "learning_rate": 8.96201653840788e-06,
      "loss": 1.5647,
      "step": 871
    },
    {
      "epoch": 0.33,
      "learning_rate": 8.825682158533554e-06,
      "loss": 1.5302,
      "step": 872
    },
    {
      "epoch": 0.33,
      "learning_rate": 8.690344846391729e-06,
      "loss": 1.3084,
      "step": 873
    },
    {
      "epoch": 0.33,
      "learning_rate": 8.55600608200956e-06,
      "loss": 1.4081,
      "step": 874
    },
    {
      "epoch": 0.33,
      "learning_rate": 8.422667334494249e-06,
      "loss": 1.4316,
      "step": 875
    },
    {
      "epoch": 0.33,
      "learning_rate": 8.290330062017016e-06,
      "loss": 1.5,
      "step": 876
    },
    {
      "epoch": 0.33,
      "learning_rate": 8.158995711797002e-06,
      "loss": 1.5807,
      "step": 877
    },
    {
      "epoch": 0.33,
      "learning_rate": 8.02866572008566e-06,
      "loss": 1.4631,
      "step": 878
    },
    {
      "epoch": 0.33,
      "learning_rate": 7.899341512150894e-06,
      "loss": 1.6158,
      "step": 879
    },
    {
      "epoch": 0.33,
      "learning_rate": 7.771024502261526e-06,
      "loss": 1.6828,
      "step": 880
    },
    {
      "epoch": 0.33,
      "learning_rate": 7.643716093671827e-06,
      "loss": 1.806,
      "step": 881
    },
    {
      "epoch": 0.33,
      "learning_rate": 7.51741767860612e-06,
      "loss": 1.4556,
      "step": 882
    },
    {
      "epoch": 0.33,
      "learning_rate": 7.392130638243666e-06,
      "loss": 1.7357,
      "step": 883
    },
    {
      "epoch": 0.33,
      "learning_rate": 7.267856342703461e-06,
      "loss": 1.5677,
      "step": 884
    },
    {
      "epoch": 0.33,
      "learning_rate": 7.144596151029303e-06,
      "loss": 1.5767,
      "step": 885
    },
    {
      "epoch": 0.33,
      "learning_rate": 7.022351411174866e-06,
      "loss": 1.483,
      "step": 886
    },
    {
      "epoch": 0.33,
      "learning_rate": 6.901123459989067e-06,
      "loss": 1.5725,
      "step": 887
    },
    {
      "epoch": 0.34,
      "learning_rate": 6.780913623201346e-06,
      "loss": 1.648,
      "step": 888
    },
    {
      "epoch": 0.34,
      "learning_rate": 6.661723215407223e-06,
      "loss": 1.3089,
      "step": 889
    },
    {
      "epoch": 0.34,
      "learning_rate": 6.543553540053926e-06,
      "loss": 1.3673,
      "step": 890
    },
    {
      "epoch": 0.34,
      "learning_rate": 6.426405889426046e-06,
      "loss": 1.7906,
      "step": 891
    },
    {
      "epoch": 0.34,
      "learning_rate": 6.310281544631546e-06,
      "loss": 1.4324,
      "step": 892
    },
    {
      "epoch": 0.34,
      "learning_rate": 6.195181775587655e-06,
      "loss": 1.5023,
      "step": 893
    },
    {
      "epoch": 0.34,
      "learning_rate": 6.081107841007006e-06,
      "loss": 1.3731,
      "step": 894
    },
    {
      "epoch": 0.34,
      "learning_rate": 5.968060988383883e-06,
      "loss": 1.3659,
      "step": 895
    },
    {
      "epoch": 0.34,
      "learning_rate": 5.856042453980526e-06,
      "loss": 1.4401,
      "step": 896
    },
    {
      "epoch": 0.34,
      "learning_rate": 5.745053462813699e-06,
      "loss": 1.2283,
      "step": 897
    },
    {
      "epoch": 0.34,
      "learning_rate": 5.63509522864123e-06,
      "loss": 1.6408,
      "step": 898
    },
    {
      "epoch": 0.34,
      "learning_rate": 5.5261689539487515e-06,
      "loss": 1.6938,
      "step": 899
    },
    {
      "epoch": 0.34,
      "learning_rate": 5.418275829936537e-06,
      "loss": 1.4477,
      "step": 900
    },
    {
      "epoch": 0.34,
      "learning_rate": 5.3114170365065164e-06,
      "loss": 1.5634,
      "step": 901
    },
    {
      "epoch": 0.34,
      "learning_rate": 5.205593742249326e-06,
      "loss": 1.4759,
      "step": 902
    },
    {
      "epoch": 0.34,
      "learning_rate": 5.100807104431571e-06,
      "loss": 1.6343,
      "step": 903
    },
    {
      "epoch": 0.34,
      "learning_rate": 4.997058268983135e-06,
      "loss": 1.5719,
      "step": 904
    },
    {
      "epoch": 0.34,
      "learning_rate": 4.8943483704846475e-06,
      "loss": 1.7144,
      "step": 905
    },
    {
      "epoch": 0.34,
      "learning_rate": 4.792678532155115e-06,
      "loss": 1.4101,
      "step": 906
    },
    {
      "epoch": 0.34,
      "learning_rate": 4.692049865839565e-06,
      "loss": 1.4293,
      "step": 907
    },
    {
      "epoch": 0.34,
      "learning_rate": 4.592463471997022e-06,
      "loss": 1.4332,
      "step": 908
    },
    {
      "epoch": 0.34,
      "learning_rate": 4.493920439688315e-06,
      "loss": 1.5606,
      "step": 909
    },
    {
      "epoch": 0.34,
      "learning_rate": 4.3964218465642355e-06,
      "loss": 1.5196,
      "step": 910
    },
    {
      "epoch": 0.34,
      "learning_rate": 4.299968758853812e-06,
      "loss": 1.7813,
      "step": 911
    },
    {
      "epoch": 0.34,
      "learning_rate": 4.204562231352516e-06,
      "loss": 1.5976,
      "step": 912
    },
    {
      "epoch": 0.34,
      "learning_rate": 4.1102033074108985e-06,
      "loss": 1.731,
      "step": 913
    },
    {
      "epoch": 0.34,
      "learning_rate": 4.016893018922996e-06,
      "loss": 1.4989,
      "step": 914
    },
    {
      "epoch": 0.35,
      "learning_rate": 3.924632386315186e-06,
      "loss": 1.4378,
      "step": 915
    },
    {
      "epoch": 0.35,
      "learning_rate": 3.83342241853496e-06,
      "loss": 1.3466,
      "step": 916
    },
    {
      "epoch": 0.35,
      "learning_rate": 3.7432641130399236e-06,
      "loss": 1.4538,
      "step": 917
    },
    {
      "epoch": 0.35,
      "learning_rate": 3.6541584557868604e-06,
      "loss": 1.5828,
      "step": 918
    },
    {
      "epoch": 0.35,
      "learning_rate": 3.5661064212209493e-06,
      "loss": 1.6145,
      "step": 919
    },
    {
      "epoch": 0.35,
      "learning_rate": 3.4791089722651436e-06,
      "loss": 1.6774,
      "step": 920
    },
    {
      "epoch": 0.35,
      "learning_rate": 3.393167060309588e-06,
      "loss": 1.4517,
      "step": 921
    },
    {
      "epoch": 0.35,
      "learning_rate": 3.3082816252012926e-06,
      "loss": 1.7202,
      "step": 922
    },
    {
      "epoch": 0.35,
      "learning_rate": 3.2244535952337562e-06,
      "loss": 1.6624,
      "step": 923
    },
    {
      "epoch": 0.35,
      "learning_rate": 3.1416838871368924e-06,
      "loss": 1.4744,
      "step": 924
    },
    {
      "epoch": 0.35,
      "learning_rate": 3.059973406066963e-06,
      "loss": 1.4983,
      "step": 925
    },
    {
      "epoch": 0.35,
      "learning_rate": 2.9793230455966937e-06,
      "loss": 1.6633,
      "step": 926
    },
    {
      "epoch": 0.35,
      "learning_rate": 2.8997336877055194e-06,
      "loss": 1.7303,
      "step": 927
    },
    {
      "epoch": 0.35,
      "learning_rate": 2.821206202769899e-06,
      "loss": 1.6822,
      "step": 928
    },
    {
      "epoch": 0.35,
      "learning_rate": 2.743741449553827e-06,
      "loss": 1.4701,
      "step": 929
    },
    {
      "epoch": 0.35,
      "learning_rate": 2.667340275199426e-06,
      "loss": 1.5291,
      "step": 930
    },
    {
      "epoch": 0.35,
      "learning_rate": 2.5920035152176892e-06,
      "loss": 1.4836,
      "step": 931
    },
    {
      "epoch": 0.35,
      "learning_rate": 2.5177319934794e-06,
      "loss": 1.5585,
      "step": 932
    },
    {
      "epoch": 0.35,
      "learning_rate": 2.44452652220597e-06,
      "loss": 1.5227,
      "step": 933
    },
    {
      "epoch": 0.35,
      "learning_rate": 2.3723879019607374e-06,
      "loss": 1.4404,
      "step": 934
    },
    {
      "epoch": 0.35,
      "learning_rate": 2.3013169216400733e-06,
      "loss": 1.602,
      "step": 935
    },
    {
      "epoch": 0.35,
      "learning_rate": 2.2313143584648423e-06,
      "loss": 1.5406,
      "step": 936
    },
    {
      "epoch": 0.35,
      "learning_rate": 2.162380977971867e-06,
      "loss": 1.4306,
      "step": 937
    },
    {
      "epoch": 0.35,
      "learning_rate": 2.0945175340055357e-06,
      "loss": 1.3999,
      "step": 938
    },
    {
      "epoch": 0.35,
      "learning_rate": 2.0277247687096155e-06,
      "loss": 1.3581,
      "step": 939
    },
    {
      "epoch": 0.35,
      "learning_rate": 1.9620034125190644e-06,
      "loss": 1.5773,
      "step": 940
    },
    {
      "epoch": 0.36,
      "learning_rate": 1.8973541841521335e-06,
      "loss": 1.3736,
      "step": 941
    },
    {
      "epoch": 0.36,
      "learning_rate": 1.8337777906023978e-06,
      "loss": 1.5649,
      "step": 942
    },
    {
      "epoch": 0.36,
      "learning_rate": 1.771274927131139e-06,
      "loss": 1.709,
      "step": 943
    },
    {
      "epoch": 0.36,
      "learning_rate": 1.7098462772596302e-06,
      "loss": 1.6171,
      "step": 944
    },
    {
      "epoch": 0.36,
      "learning_rate": 1.6494925127617634e-06,
      "loss": 1.4633,
      "step": 945
    },
    {
      "epoch": 0.36,
      "learning_rate": 1.5902142936566334e-06,
      "loss": 1.7157,
      "step": 946
    },
    {
      "epoch": 0.36,
      "learning_rate": 1.532012268201344e-06,
      "loss": 1.5205,
      "step": 947
    },
    {
      "epoch": 0.36,
      "learning_rate": 1.4748870728839347e-06,
      "loss": 1.5265,
      "step": 948
    },
    {
      "epoch": 0.36,
      "learning_rate": 1.4188393324163663e-06,
      "loss": 1.4795,
      "step": 949
    },
    {
      "epoch": 0.36,
      "learning_rate": 1.3638696597277679e-06,
      "loss": 1.4224,
      "step": 950
    },
    {
      "epoch": 0.36,
      "learning_rate": 1.3099786559576554e-06,
      "loss": 1.5683,
      "step": 951
    },
    {
      "epoch": 0.36,
      "learning_rate": 1.2571669104494256e-06,
      "loss": 1.3445,
      "step": 952
    },
    {
      "epoch": 0.36,
      "learning_rate": 1.2054350007438707e-06,
      "loss": 1.3331,
      "step": 953
    },
    {
      "epoch": 0.36,
      "learning_rate": 1.1547834925728528e-06,
      "loss": 1.4435,
      "step": 954
    },
    {
      "epoch": 0.36,
      "learning_rate": 1.1052129398531507e-06,
      "loss": 1.6942,
      "step": 955
    },
    {
      "epoch": 0.36,
      "learning_rate": 1.0567238846803996e-06,
      "loss": 1.471,
      "step": 956
    },
    {
      "epoch": 0.36,
      "learning_rate": 1.0093168573231393e-06,
      "loss": 1.3496,
      "step": 957
    },
    {
      "epoch": 0.36,
      "learning_rate": 9.62992376217009e-07,
      "loss": 1.4363,
      "step": 958
    },
    {
      "epoch": 0.36,
      "learning_rate": 9.177509479591173e-07,
      "loss": 1.2465,
      "step": 959
    },
    {
      "epoch": 0.36,
      "learning_rate": 8.735930673024806e-07,
      "loss": 1.4027,
      "step": 960
    },
    {
      "epoch": 0.36,
      "learning_rate": 8.305192171506049e-07,
      "loss": 1.6351,
      "step": 961
    },
    {
      "epoch": 0.36,
      "learning_rate": 7.885298685522235e-07,
      "loss": 1.4707,
      "step": 962
    },
    {
      "epoch": 0.36,
      "learning_rate": 7.476254806961014e-07,
      "loss": 1.6195,
      "step": 963
    },
    {
      "epoch": 0.36,
      "learning_rate": 7.078065009060941e-07,
      "loss": 1.5442,
      "step": 964
    },
    {
      "epoch": 0.36,
      "learning_rate": 6.690733646361857e-07,
      "loss": 1.492,
      "step": 965
    },
    {
      "epoch": 0.36,
      "learning_rate": 6.314264954657256e-07,
      "loss": 1.479,
      "step": 966
    },
    {
      "epoch": 0.36,
      "learning_rate": 5.948663050948767e-07,
      "loss": 1.6958,
      "step": 967
    },
    {
      "epoch": 0.37,
      "learning_rate": 5.593931933399854e-07,
      "loss": 1.6097,
      "step": 968
    },
    {
      "epoch": 0.37,
      "learning_rate": 5.250075481293526e-07,
      "loss": 1.5101,
      "step": 969
    },
    {
      "epoch": 0.37,
      "learning_rate": 4.917097454988584e-07,
      "loss": 1.5687,
      "step": 970
    },
    {
      "epoch": 0.37,
      "learning_rate": 4.5950014958795475e-07,
      "loss": 1.4106,
      "step": 971
    },
    {
      "epoch": 0.37,
      "learning_rate": 4.2837911263562404e-07,
      "loss": 1.4372,
      "step": 972
    },
    {
      "epoch": 0.37,
      "learning_rate": 3.983469749765267e-07,
      "loss": 1.5331,
      "step": 973
    },
    {
      "epoch": 0.37,
      "learning_rate": 3.694040650373154e-07,
      "loss": 1.816,
      "step": 974
    },
    {
      "epoch": 0.37,
      "learning_rate": 3.415506993330153e-07,
      "loss": 1.5552,
      "step": 975
    },
    {
      "epoch": 0.37,
      "learning_rate": 3.1478718246357173e-07,
      "loss": 1.4554,
      "step": 976
    },
    {
      "epoch": 0.37,
      "learning_rate": 2.891138071105193e-07,
      "loss": 1.4739,
      "step": 977
    },
    {
      "epoch": 0.37,
      "learning_rate": 2.645308540337843e-07,
      "loss": 1.5654,
      "step": 978
    },
    {
      "epoch": 0.37,
      "learning_rate": 2.410385920685765e-07,
      "loss": 1.4487,
      "step": 979
    },
    {
      "epoch": 0.37,
      "learning_rate": 2.1863727812254653e-07,
      "loss": 1.4449,
      "step": 980
    },
    {
      "epoch": 0.37,
      "learning_rate": 1.973271571728441e-07,
      "loss": 1.517,
      "step": 981
    },
    {
      "epoch": 0.37,
      "learning_rate": 1.7710846226355328e-07,
      "loss": 1.5971,
      "step": 982
    },
    {
      "epoch": 0.37,
      "learning_rate": 1.5798141450307225e-07,
      "loss": 1.6616,
      "step": 983
    },
    {
      "epoch": 0.37,
      "learning_rate": 1.3994622306173765e-07,
      "loss": 1.3977,
      "step": 984
    },
    {
      "epoch": 0.37,
      "learning_rate": 1.230030851695263e-07,
      "loss": 1.5483,
      "step": 985
    },
    {
      "epoch": 0.37,
      "learning_rate": 1.0715218611384581e-07,
      "loss": 1.9048,
      "step": 986
    },
    {
      "epoch": 0.37,
      "learning_rate": 9.239369923762508e-08,
      "loss": 1.5007,
      "step": 987
    },
    {
      "epoch": 0.37,
      "learning_rate": 7.872778593728258e-08,
      "loss": 1.5753,
      "step": 988
    },
    {
      "epoch": 0.37,
      "learning_rate": 6.615459566108317e-08,
      "loss": 1.4905,
      "step": 989
    },
    {
      "epoch": 0.37,
      "learning_rate": 5.467426590739511e-08,
      "loss": 1.525,
      "step": 990
    },
    {
      "epoch": 0.37,
      "learning_rate": 4.428692222329112e-08,
      "loss": 1.5879,
      "step": 991
    },
    {
      "epoch": 0.37,
      "learning_rate": 3.499267820307184e-08,
      "loss": 1.5155,
      "step": 992
    },
    {
      "epoch": 0.37,
      "learning_rate": 2.6791635487122268e-08,
      "loss": 1.4999,
      "step": 993
    },
    {
      "epoch": 0.38,
      "learning_rate": 1.9683883760723832e-08,
      "loss": 1.5192,
      "step": 994
    },
    {
      "epoch": 0.38,
      "learning_rate": 1.3669500753099585e-08,
      "loss": 1.6683,
      "step": 995
    },
    {
      "epoch": 0.38,
      "learning_rate": 8.748552236603757e-09,
      "loss": 1.6939,
      "step": 996
    },
    {
      "epoch": 0.38,
      "learning_rate": 4.921092025911289e-09,
      "loss": 1.4738,
      "step": 997
    },
    {
      "epoch": 0.38,
      "learning_rate": 2.187161977540431e-09,
      "loss": 1.3242,
      "step": 998
    },
    {
      "epoch": 0.38,
      "learning_rate": 5.467919892865325e-10,
      "loss": 1.4856,
      "step": 999
    },
    {
      "epoch": 0.38,
      "learning_rate": 0.0,
      "loss": 1.2842,
      "step": 1000
    }
  ],
  "max_steps": 1000,
  "num_train_epochs": 1,
  "total_flos": 4.212995477615309e+16,
  "trial_name": null,
  "trial_params": null
}