{
  "best_metric": 0.08992622047662735,
  "best_model_checkpoint": "./phishing-email_sender-detection/checkpoint-2733",
  "epoch": 1.0,
  "global_step": 2733,
  "is_hyper_param_search": false,
  "is_local_process_zero": true,
  "is_world_process_zero": true,
  "log_history": [
    {
      "epoch": 0.0,
      "learning_rate": 1.0000000000000002e-06,
      "loss": 0.6822,
      "step": 1
    },
    {
      "epoch": 0.0,
      "learning_rate": 2.0000000000000003e-06,
      "loss": 0.6864,
      "step": 2
    },
    {
      "epoch": 0.0,
      "learning_rate": 3e-06,
      "loss": 0.6799,
      "step": 3
    },
    {
      "epoch": 0.0,
      "learning_rate": 4.000000000000001e-06,
      "loss": 0.7138,
      "step": 4
    },
    {
      "epoch": 0.0,
      "learning_rate": 5e-06,
      "loss": 0.6736,
      "step": 5
    },
    {
      "epoch": 0.0,
      "learning_rate": 6e-06,
      "loss": 0.7144,
      "step": 6
    },
    {
      "epoch": 0.0,
      "learning_rate": 7.000000000000001e-06,
      "loss": 0.708,
      "step": 7
    },
    {
      "epoch": 0.0,
      "learning_rate": 8.000000000000001e-06,
      "loss": 0.6725,
      "step": 8
    },
    {
      "epoch": 0.0,
      "learning_rate": 9e-06,
      "loss": 0.695,
      "step": 9
    },
    {
      "epoch": 0.0,
      "learning_rate": 1e-05,
      "loss": 0.6955,
      "step": 10
    },
    {
      "epoch": 0.0,
      "learning_rate": 1.1000000000000001e-05,
      "loss": 0.6922,
      "step": 11
    },
    {
      "epoch": 0.0,
      "learning_rate": 1.2e-05,
      "loss": 0.6767,
      "step": 12
    },
    {
      "epoch": 0.0,
      "learning_rate": 1.3000000000000001e-05,
      "loss": 0.6871,
      "step": 13
    },
    {
      "epoch": 0.01,
      "learning_rate": 1.4000000000000001e-05,
      "loss": 0.6951,
      "step": 14
    },
    {
      "epoch": 0.01,
      "learning_rate": 1.5e-05,
      "loss": 0.6666,
      "step": 15
    },
    {
      "epoch": 0.01,
      "learning_rate": 1.6000000000000003e-05,
      "loss": 0.6821,
      "step": 16
    },
    {
      "epoch": 0.01,
      "learning_rate": 1.7000000000000003e-05,
      "loss": 0.6521,
      "step": 17
    },
    {
      "epoch": 0.01,
      "learning_rate": 1.8e-05,
      "loss": 0.6854,
      "step": 18
    },
    {
      "epoch": 0.01,
      "learning_rate": 1.9e-05,
      "loss": 0.6891,
      "step": 19
    },
    {
      "epoch": 0.01,
      "learning_rate": 2e-05,
      "loss": 0.6533,
      "step": 20
    },
    {
      "epoch": 0.01,
      "learning_rate": 2.1e-05,
      "loss": 0.6493,
      "step": 21
    },
    {
      "epoch": 0.01,
      "learning_rate": 2.2000000000000003e-05,
      "loss": 0.6055,
      "step": 22
    },
    {
      "epoch": 0.01,
      "learning_rate": 2.3000000000000003e-05,
      "loss": 0.611,
      "step": 23
    },
    {
      "epoch": 0.01,
      "learning_rate": 2.4e-05,
      "loss": 0.644,
      "step": 24
    },
    {
      "epoch": 0.01,
      "learning_rate": 2.5e-05,
      "loss": 0.5345,
      "step": 25
    },
    {
      "epoch": 0.01,
      "learning_rate": 2.6000000000000002e-05,
      "loss": 0.5509,
      "step": 26
    },
    {
      "epoch": 0.01,
      "learning_rate": 2.7000000000000002e-05,
      "loss": 0.6089,
      "step": 27
    },
    {
      "epoch": 0.01,
      "learning_rate": 2.8000000000000003e-05,
      "loss": 0.5878,
      "step": 28
    },
    {
      "epoch": 0.01,
      "learning_rate": 2.9e-05,
      "loss": 0.4539,
      "step": 29
    },
    {
      "epoch": 0.01,
      "learning_rate": 3e-05,
      "loss": 0.5623,
      "step": 30
    },
    {
      "epoch": 0.01,
      "learning_rate": 3.1e-05,
      "loss": 0.5357,
      "step": 31
    },
    {
      "epoch": 0.01,
      "learning_rate": 3.2000000000000005e-05,
      "loss": 0.4834,
      "step": 32
    },
    {
      "epoch": 0.01,
      "learning_rate": 3.3e-05,
      "loss": 0.8397,
      "step": 33
    },
    {
      "epoch": 0.01,
      "learning_rate": 3.4000000000000007e-05,
      "loss": 0.5306,
      "step": 34
    },
    {
      "epoch": 0.01,
      "learning_rate": 3.5e-05,
      "loss": 0.5296,
      "step": 35
    },
    {
      "epoch": 0.01,
      "learning_rate": 3.6e-05,
      "loss": 0.4745,
      "step": 36
    },
    {
      "epoch": 0.01,
      "learning_rate": 3.7e-05,
      "loss": 0.5718,
      "step": 37
    },
    {
      "epoch": 0.01,
      "learning_rate": 3.8e-05,
      "loss": 0.4702,
      "step": 38
    },
    {
      "epoch": 0.01,
      "learning_rate": 3.9000000000000006e-05,
      "loss": 0.4885,
      "step": 39
    },
    {
      "epoch": 0.01,
      "learning_rate": 4e-05,
      "loss": 0.4,
      "step": 40
    },
    {
      "epoch": 0.02,
      "learning_rate": 4.1e-05,
      "loss": 0.4456,
      "step": 41
    },
    {
      "epoch": 0.02,
      "learning_rate": 4.2e-05,
      "loss": 0.3533,
      "step": 42
    },
    {
      "epoch": 0.02,
      "learning_rate": 4.3e-05,
      "loss": 0.3855,
      "step": 43
    },
    {
      "epoch": 0.02,
      "learning_rate": 4.4000000000000006e-05,
      "loss": 0.3514,
      "step": 44
    },
    {
      "epoch": 0.02,
      "learning_rate": 4.5e-05,
      "loss": 0.511,
      "step": 45
    },
    {
      "epoch": 0.02,
      "learning_rate": 4.600000000000001e-05,
      "loss": 0.5808,
      "step": 46
    },
    {
      "epoch": 0.02,
      "learning_rate": 4.7e-05,
      "loss": 0.4821,
      "step": 47
    },
    {
      "epoch": 0.02,
      "learning_rate": 4.8e-05,
      "loss": 0.2828,
      "step": 48
    },
    {
      "epoch": 0.02,
      "learning_rate": 4.9e-05,
      "loss": 0.5577,
      "step": 49
    },
    {
      "epoch": 0.02,
      "learning_rate": 5e-05,
      "loss": 0.4195,
      "step": 50
    },
    {
      "epoch": 0.02,
      "learning_rate": 4.999076809453471e-05,
      "loss": 0.2299,
      "step": 51
    },
    {
      "epoch": 0.02,
      "learning_rate": 4.998153618906943e-05,
      "loss": 0.5267,
      "step": 52
    },
    {
      "epoch": 0.02,
      "learning_rate": 4.997230428360414e-05,
      "loss": 0.3875,
      "step": 53
    },
    {
      "epoch": 0.02,
      "learning_rate": 4.9963072378138845e-05,
      "loss": 0.6143,
      "step": 54
    },
    {
      "epoch": 0.02,
      "learning_rate": 4.995384047267356e-05,
      "loss": 0.4869,
      "step": 55
    },
    {
      "epoch": 0.02,
      "learning_rate": 4.994460856720827e-05,
      "loss": 0.3315,
      "step": 56
    },
    {
      "epoch": 0.02,
      "learning_rate": 4.993537666174299e-05,
      "loss": 0.1522,
      "step": 57
    },
    {
      "epoch": 0.02,
      "learning_rate": 4.99261447562777e-05,
      "loss": 0.2578,
      "step": 58
    },
    {
      "epoch": 0.02,
      "learning_rate": 4.991691285081241e-05,
      "loss": 0.3901,
      "step": 59
    },
    {
      "epoch": 0.02,
      "learning_rate": 4.990768094534712e-05,
      "loss": 0.5306,
      "step": 60
    },
    {
      "epoch": 0.02,
      "learning_rate": 4.989844903988183e-05,
      "loss": 0.5296,
      "step": 61
    },
    {
      "epoch": 0.02,
      "learning_rate": 4.988921713441654e-05,
      "loss": 0.1696,
      "step": 62
    },
    {
      "epoch": 0.02,
      "learning_rate": 4.987998522895126e-05,
      "loss": 0.3566,
      "step": 63
    },
    {
      "epoch": 0.02,
      "learning_rate": 4.987075332348597e-05,
      "loss": 0.6988,
      "step": 64
    },
    {
      "epoch": 0.02,
      "learning_rate": 4.986152141802068e-05,
      "loss": 0.2724,
      "step": 65
    },
    {
      "epoch": 0.02,
      "learning_rate": 4.985228951255539e-05,
      "loss": 0.4354,
      "step": 66
    },
    {
      "epoch": 0.02,
      "learning_rate": 4.98430576070901e-05,
      "loss": 0.6792,
      "step": 67
    },
    {
      "epoch": 0.02,
      "learning_rate": 4.983382570162482e-05,
      "loss": 0.2795,
      "step": 68
    },
    {
      "epoch": 0.03,
      "learning_rate": 4.982459379615953e-05,
      "loss": 0.4571,
      "step": 69
    },
    {
      "epoch": 0.03,
      "learning_rate": 4.981536189069424e-05,
      "loss": 0.3376,
      "step": 70
    },
    {
      "epoch": 0.03,
      "learning_rate": 4.980612998522896e-05,
      "loss": 0.4101,
      "step": 71
    },
    {
      "epoch": 0.03,
      "learning_rate": 4.979689807976367e-05,
      "loss": 0.2786,
      "step": 72
    },
    {
      "epoch": 0.03,
      "learning_rate": 4.978766617429837e-05,
      "loss": 0.8762,
      "step": 73
    },
    {
      "epoch": 0.03,
      "learning_rate": 4.977843426883309e-05,
      "loss": 0.2465,
      "step": 74
    },
    {
      "epoch": 0.03,
      "learning_rate": 4.97692023633678e-05,
      "loss": 0.4012,
      "step": 75
    },
    {
      "epoch": 0.03,
      "learning_rate": 4.975997045790252e-05,
      "loss": 0.3297,
      "step": 76
    },
    {
      "epoch": 0.03,
      "learning_rate": 4.975073855243723e-05,
      "loss": 0.1251,
      "step": 77
    },
    {
      "epoch": 0.03,
      "learning_rate": 4.974150664697194e-05,
      "loss": 0.2363,
      "step": 78
    },
    {
      "epoch": 0.03,
      "learning_rate": 4.973227474150665e-05,
      "loss": 0.2676,
      "step": 79
    },
    {
      "epoch": 0.03,
      "learning_rate": 4.972304283604136e-05,
      "loss": 0.3833,
      "step": 80
    },
    {
      "epoch": 0.03,
      "learning_rate": 4.971381093057607e-05,
      "loss": 0.4222,
      "step": 81
    },
    {
      "epoch": 0.03,
      "learning_rate": 4.970457902511079e-05,
      "loss": 0.5522,
      "step": 82
    },
    {
      "epoch": 0.03,
      "learning_rate": 4.96953471196455e-05,
      "loss": 0.2485,
      "step": 83
    },
    {
      "epoch": 0.03,
      "learning_rate": 4.968611521418021e-05,
      "loss": 0.3908,
      "step": 84
    },
    {
      "epoch": 0.03,
      "learning_rate": 4.967688330871492e-05,
      "loss": 0.542,
      "step": 85
    },
    {
      "epoch": 0.03,
      "learning_rate": 4.966765140324963e-05,
      "loss": 0.6979,
      "step": 86
    },
    {
      "epoch": 0.03,
      "learning_rate": 4.965841949778435e-05,
      "loss": 0.4682,
      "step": 87
    },
    {
      "epoch": 0.03,
      "learning_rate": 4.964918759231906e-05,
      "loss": 0.3517,
      "step": 88
    },
    {
      "epoch": 0.03,
      "learning_rate": 4.963995568685377e-05,
      "loss": 0.309,
      "step": 89
    },
    {
      "epoch": 0.03,
      "learning_rate": 4.9630723781388485e-05,
      "loss": 0.1658,
      "step": 90
    },
    {
      "epoch": 0.03,
      "learning_rate": 4.9621491875923196e-05,
      "loss": 0.2953,
      "step": 91
    },
    {
      "epoch": 0.03,
      "learning_rate": 4.96122599704579e-05,
      "loss": 0.3668,
      "step": 92
    },
    {
      "epoch": 0.03,
      "learning_rate": 4.960302806499262e-05,
      "loss": 0.1059,
      "step": 93
    },
    {
      "epoch": 0.03,
      "learning_rate": 4.959379615952733e-05,
      "loss": 0.767,
      "step": 94
    },
    {
      "epoch": 0.03,
      "learning_rate": 4.958456425406204e-05,
      "loss": 0.6302,
      "step": 95
    },
    {
      "epoch": 0.04,
      "learning_rate": 4.9575332348596755e-05,
      "loss": 0.3325,
      "step": 96
    },
    {
      "epoch": 0.04,
      "learning_rate": 4.9566100443131466e-05,
      "loss": 0.2238,
      "step": 97
    },
    {
      "epoch": 0.04,
      "learning_rate": 4.9556868537666177e-05,
      "loss": 0.4277,
      "step": 98
    },
    {
      "epoch": 0.04,
      "learning_rate": 4.954763663220089e-05,
      "loss": 0.261,
      "step": 99
    },
    {
      "epoch": 0.04,
      "learning_rate": 4.95384047267356e-05,
      "loss": 0.3343,
      "step": 100
    },
    {
      "epoch": 0.04,
      "learning_rate": 4.9529172821270315e-05,
      "loss": 0.3266,
      "step": 101
    },
    {
      "epoch": 0.04,
      "learning_rate": 4.9519940915805025e-05,
      "loss": 0.3277,
      "step": 102
    },
    {
      "epoch": 0.04,
      "learning_rate": 4.9510709010339736e-05,
      "loss": 0.2546,
      "step": 103
    },
    {
      "epoch": 0.04,
      "learning_rate": 4.9501477104874447e-05,
      "loss": 0.3059,
      "step": 104
    },
    {
      "epoch": 0.04,
      "learning_rate": 4.949224519940916e-05,
      "loss": 0.5196,
      "step": 105
    },
    {
      "epoch": 0.04,
      "learning_rate": 4.9483013293943874e-05,
      "loss": 0.2356,
      "step": 106
    },
    {
      "epoch": 0.04,
      "learning_rate": 4.9473781388478585e-05,
      "loss": 0.3281,
      "step": 107
    },
    {
      "epoch": 0.04,
      "learning_rate": 4.9464549483013296e-05,
      "loss": 0.3736,
      "step": 108
    },
    {
      "epoch": 0.04,
      "learning_rate": 4.945531757754801e-05,
      "loss": 0.2189,
      "step": 109
    },
    {
      "epoch": 0.04,
      "learning_rate": 4.9446085672082723e-05,
      "loss": 0.5514,
      "step": 110
    },
    {
      "epoch": 0.04,
      "learning_rate": 4.943685376661743e-05,
      "loss": 0.2095,
      "step": 111
    },
    {
      "epoch": 0.04,
      "learning_rate": 4.9427621861152145e-05,
      "loss": 0.1746,
      "step": 112
    },
    {
      "epoch": 0.04,
      "learning_rate": 4.9418389955686855e-05,
      "loss": 0.2804,
      "step": 113
    },
    {
      "epoch": 0.04,
      "learning_rate": 4.9409158050221566e-05,
      "loss": 0.2254,
      "step": 114
    },
    {
      "epoch": 0.04,
      "learning_rate": 4.939992614475628e-05,
      "loss": 0.3178,
      "step": 115
    },
    {
      "epoch": 0.04,
      "learning_rate": 4.9390694239290994e-05,
      "loss": 0.1731,
      "step": 116
    },
    {
      "epoch": 0.04,
      "learning_rate": 4.9381462333825704e-05,
      "loss": 0.1595,
      "step": 117
    },
    {
      "epoch": 0.04,
      "learning_rate": 4.9372230428360415e-05,
      "loss": 0.3126,
      "step": 118
    },
    {
      "epoch": 0.04,
      "learning_rate": 4.9362998522895125e-05,
      "loss": 0.2202,
      "step": 119
    },
    {
      "epoch": 0.04,
      "learning_rate": 4.935376661742984e-05,
      "loss": 0.2444,
      "step": 120
    },
    {
      "epoch": 0.04,
      "learning_rate": 4.934453471196455e-05,
      "loss": 0.4146,
      "step": 121
    },
    {
      "epoch": 0.04,
      "learning_rate": 4.9335302806499264e-05,
      "loss": 0.7269,
      "step": 122
    },
    {
      "epoch": 0.05,
      "learning_rate": 4.9326070901033974e-05,
      "loss": 0.3328,
      "step": 123
    },
    {
      "epoch": 0.05,
      "learning_rate": 4.9316838995568685e-05,
      "loss": 0.1629,
      "step": 124
    },
    {
      "epoch": 0.05,
      "learning_rate": 4.9307607090103395e-05,
      "loss": 0.1924,
      "step": 125
    },
    {
      "epoch": 0.05,
      "learning_rate": 4.929837518463811e-05,
      "loss": 0.3693,
      "step": 126
    },
    {
      "epoch": 0.05,
      "learning_rate": 4.928914327917282e-05,
      "loss": 0.2883,
      "step": 127
    },
    {
      "epoch": 0.05,
      "learning_rate": 4.927991137370754e-05,
      "loss": 0.2518,
      "step": 128
    },
    {
      "epoch": 0.05,
      "learning_rate": 4.927067946824225e-05,
      "loss": 0.3868,
      "step": 129
    },
    {
      "epoch": 0.05,
      "learning_rate": 4.9261447562776955e-05,
      "loss": 0.2361,
      "step": 130
    },
    {
      "epoch": 0.05,
      "learning_rate": 4.925221565731167e-05,
      "loss": 0.2834,
      "step": 131
    },
    {
      "epoch": 0.05,
      "learning_rate": 4.924298375184638e-05,
      "loss": 0.1881,
      "step": 132
    },
    {
      "epoch": 0.05,
      "learning_rate": 4.923375184638109e-05,
      "loss": 0.3652,
      "step": 133
    },
    {
      "epoch": 0.05,
      "learning_rate": 4.922451994091581e-05,
      "loss": 0.0794,
      "step": 134
    },
    {
      "epoch": 0.05,
      "learning_rate": 4.921528803545052e-05,
      "loss": 0.3755,
      "step": 135
    },
    {
      "epoch": 0.05,
      "learning_rate": 4.920605612998523e-05,
      "loss": 0.2346,
      "step": 136
    },
    {
      "epoch": 0.05,
      "learning_rate": 4.919682422451994e-05,
      "loss": 0.1665,
      "step": 137
    },
    {
      "epoch": 0.05,
      "learning_rate": 4.918759231905465e-05,
      "loss": 0.1967,
      "step": 138
    },
    {
      "epoch": 0.05,
      "learning_rate": 4.917836041358937e-05,
      "loss": 0.2299,
      "step": 139
    },
    {
      "epoch": 0.05,
      "learning_rate": 4.916912850812408e-05,
      "loss": 0.1734,
      "step": 140
    },
    {
      "epoch": 0.05,
      "learning_rate": 4.915989660265879e-05,
      "loss": 0.196,
      "step": 141
    },
    {
      "epoch": 0.05,
      "learning_rate": 4.91506646971935e-05,
      "loss": 0.3994,
      "step": 142
    },
    {
      "epoch": 0.05,
      "learning_rate": 4.914143279172821e-05,
      "loss": 0.1139,
      "step": 143
    },
    {
      "epoch": 0.05,
      "learning_rate": 4.913220088626292e-05,
      "loss": 0.1924,
      "step": 144
    },
    {
      "epoch": 0.05,
      "learning_rate": 4.912296898079764e-05,
      "loss": 0.2402,
      "step": 145
    },
    {
      "epoch": 0.05,
      "learning_rate": 4.911373707533235e-05,
      "loss": 0.1419,
      "step": 146
    },
    {
      "epoch": 0.05,
      "learning_rate": 4.910450516986707e-05,
      "loss": 0.1935,
      "step": 147
    },
    {
      "epoch": 0.05,
      "learning_rate": 4.909527326440178e-05,
      "loss": 0.2484,
      "step": 148
    },
    {
      "epoch": 0.05,
      "learning_rate": 4.908604135893648e-05,
      "loss": 0.3623,
      "step": 149
    },
    {
      "epoch": 0.05,
      "learning_rate": 4.90768094534712e-05,
      "loss": 0.0689,
      "step": 150
    },
    {
      "epoch": 0.06,
      "learning_rate": 4.906757754800591e-05,
      "loss": 0.1651,
      "step": 151
    },
    {
      "epoch": 0.06,
      "learning_rate": 4.905834564254062e-05,
      "loss": 0.1806,
      "step": 152
    },
    {
      "epoch": 0.06,
      "learning_rate": 4.904911373707534e-05,
      "loss": 0.4273,
      "step": 153
    },
    {
      "epoch": 0.06,
      "learning_rate": 4.903988183161005e-05,
      "loss": 0.5241,
      "step": 154
    },
    {
      "epoch": 0.06,
      "learning_rate": 4.903064992614476e-05,
      "loss": 0.1556,
      "step": 155
    },
    {
      "epoch": 0.06,
      "learning_rate": 4.902141802067947e-05,
      "loss": 0.1175,
      "step": 156
    },
    {
      "epoch": 0.06,
      "learning_rate": 4.901218611521418e-05,
      "loss": 0.1229,
      "step": 157
    },
    {
      "epoch": 0.06,
      "learning_rate": 4.90029542097489e-05,
      "loss": 0.2728,
      "step": 158
    },
    {
      "epoch": 0.06,
      "learning_rate": 4.899372230428361e-05,
      "loss": 0.0694,
      "step": 159
    },
    {
      "epoch": 0.06,
      "learning_rate": 4.898449039881832e-05,
      "loss": 0.6585,
      "step": 160
    },
    {
      "epoch": 0.06,
      "learning_rate": 4.897525849335303e-05,
      "loss": 0.3303,
      "step": 161
    },
    {
      "epoch": 0.06,
      "learning_rate": 4.896602658788774e-05,
      "loss": 0.2365,
      "step": 162
    },
    {
      "epoch": 0.06,
      "learning_rate": 4.895679468242245e-05,
      "loss": 0.4375,
      "step": 163
    },
    {
      "epoch": 0.06,
      "learning_rate": 4.894756277695717e-05,
      "loss": 0.3717,
      "step": 164
    },
    {
      "epoch": 0.06,
      "learning_rate": 4.893833087149188e-05,
      "loss": 0.1246,
      "step": 165
    },
    {
      "epoch": 0.06,
      "learning_rate": 4.892909896602659e-05,
      "loss": 0.1015,
      "step": 166
    },
    {
      "epoch": 0.06,
      "learning_rate": 4.8919867060561306e-05,
      "loss": 0.2132,
      "step": 167
    },
    {
      "epoch": 0.06,
      "learning_rate": 4.891063515509601e-05,
      "loss": 0.3338,
      "step": 168
    },
    {
      "epoch": 0.06,
      "learning_rate": 4.890140324963073e-05,
      "loss": 0.5773,
      "step": 169
    },
    {
      "epoch": 0.06,
      "learning_rate": 4.889217134416544e-05,
      "loss": 0.1864,
      "step": 170
    },
    {
      "epoch": 0.06,
      "learning_rate": 4.888293943870015e-05,
      "loss": 0.2743,
      "step": 171
    },
    {
      "epoch": 0.06,
      "learning_rate": 4.8873707533234866e-05,
      "loss": 0.3322,
      "step": 172
    },
    {
      "epoch": 0.06,
      "learning_rate": 4.8864475627769576e-05,
      "loss": 0.1509,
      "step": 173
    },
    {
      "epoch": 0.06,
      "learning_rate": 4.885524372230429e-05,
      "loss": 0.1388,
      "step": 174
    },
    {
      "epoch": 0.06,
      "learning_rate": 4.8846011816839e-05,
      "loss": 0.2207,
      "step": 175
    },
    {
      "epoch": 0.06,
      "learning_rate": 4.883677991137371e-05,
      "loss": 0.1739,
      "step": 176
    },
    {
      "epoch": 0.06,
      "learning_rate": 4.8827548005908425e-05,
      "loss": 0.1133,
      "step": 177
    },
    {
      "epoch": 0.07,
      "learning_rate": 4.8818316100443136e-05,
      "loss": 0.6126,
      "step": 178
    },
    {
      "epoch": 0.07,
      "learning_rate": 4.8809084194977846e-05,
      "loss": 0.5063,
      "step": 179
    },
    {
      "epoch": 0.07,
      "learning_rate": 4.879985228951256e-05,
      "loss": 0.1269,
      "step": 180
    },
    {
      "epoch": 0.07,
      "learning_rate": 4.879062038404727e-05,
      "loss": 0.1424,
      "step": 181
    },
    {
      "epoch": 0.07,
      "learning_rate": 4.878138847858198e-05,
      "loss": 0.065,
      "step": 182
    },
    {
      "epoch": 0.07,
      "learning_rate": 4.8772156573116695e-05,
      "loss": 0.0509,
      "step": 183
    },
    {
      "epoch": 0.07,
      "learning_rate": 4.8762924667651406e-05,
      "loss": 0.2515,
      "step": 184
    },
    {
      "epoch": 0.07,
      "learning_rate": 4.8753692762186116e-05,
      "loss": 0.6077,
      "step": 185
    },
    {
      "epoch": 0.07,
      "learning_rate": 4.8744460856720834e-05,
      "loss": 0.0153,
      "step": 186
    },
    {
      "epoch": 0.07,
      "learning_rate": 4.873522895125554e-05,
      "loss": 0.5614,
      "step": 187
    },
    {
      "epoch": 0.07,
      "learning_rate": 4.8725997045790255e-05,
      "loss": 0.2065,
      "step": 188
    },
    {
      "epoch": 0.07,
      "learning_rate": 4.8716765140324965e-05,
      "loss": 0.1116,
      "step": 189
    },
    {
      "epoch": 0.07,
      "learning_rate": 4.8707533234859676e-05,
      "loss": 0.0205,
      "step": 190
    },
    {
      "epoch": 0.07,
      "learning_rate": 4.869830132939439e-05,
      "loss": 0.1511,
      "step": 191
    },
    {
      "epoch": 0.07,
      "learning_rate": 4.8689069423929104e-05,
      "loss": 0.1373,
      "step": 192
    },
    {
      "epoch": 0.07,
      "learning_rate": 4.8679837518463814e-05,
      "loss": 0.0549,
      "step": 193
    },
    {
      "epoch": 0.07,
      "learning_rate": 4.8670605612998525e-05,
      "loss": 0.4043,
      "step": 194
    },
    {
      "epoch": 0.07,
      "learning_rate": 4.8661373707533235e-05,
      "loss": 0.388,
      "step": 195
    },
    {
      "epoch": 0.07,
      "learning_rate": 4.8652141802067946e-05,
      "loss": 0.6158,
      "step": 196
    },
    {
      "epoch": 0.07,
      "learning_rate": 4.864290989660266e-05,
      "loss": 0.3313,
      "step": 197
    },
    {
      "epoch": 0.07,
      "learning_rate": 4.8633677991137374e-05,
      "loss": 0.6736,
      "step": 198
    },
    {
      "epoch": 0.07,
      "learning_rate": 4.8624446085672084e-05,
      "loss": 0.4404,
      "step": 199
    },
    {
      "epoch": 0.07,
      "learning_rate": 4.8615214180206795e-05,
      "loss": 0.3187,
      "step": 200
    },
    {
      "epoch": 0.07,
      "learning_rate": 4.8605982274741505e-05,
      "loss": 0.0857,
      "step": 201
    },
    {
      "epoch": 0.07,
      "learning_rate": 4.859675036927622e-05,
      "loss": 0.166,
      "step": 202
    },
    {
      "epoch": 0.07,
      "learning_rate": 4.858751846381093e-05,
      "loss": 0.1839,
      "step": 203
    },
    {
      "epoch": 0.07,
      "learning_rate": 4.8578286558345644e-05,
      "loss": 0.1136,
      "step": 204
    },
    {
      "epoch": 0.08,
      "learning_rate": 4.856905465288036e-05,
      "loss": 0.1336,
      "step": 205
    },
    {
      "epoch": 0.08,
      "learning_rate": 4.8559822747415065e-05,
      "loss": 0.0638,
      "step": 206
    },
    {
      "epoch": 0.08,
      "learning_rate": 4.855059084194978e-05,
      "loss": 0.053,
      "step": 207
    },
    {
      "epoch": 0.08,
      "learning_rate": 4.854135893648449e-05,
      "loss": 0.0611,
      "step": 208
    },
    {
      "epoch": 0.08,
      "learning_rate": 4.85321270310192e-05,
      "loss": 0.4008,
      "step": 209
    },
    {
      "epoch": 0.08,
      "learning_rate": 4.852289512555392e-05,
      "loss": 0.134,
      "step": 210
    },
    {
      "epoch": 0.08,
      "learning_rate": 4.851366322008863e-05,
      "loss": 0.1926,
      "step": 211
    },
    {
      "epoch": 0.08,
      "learning_rate": 4.850443131462334e-05,
      "loss": 0.4106,
      "step": 212
    },
    {
      "epoch": 0.08,
      "learning_rate": 4.849519940915805e-05,
      "loss": 0.165,
      "step": 213
    },
    {
      "epoch": 0.08,
      "learning_rate": 4.848596750369276e-05,
      "loss": 0.2622,
      "step": 214
    },
    {
      "epoch": 0.08,
      "learning_rate": 4.847673559822747e-05,
      "loss": 0.141,
      "step": 215
    },
    {
      "epoch": 0.08,
      "learning_rate": 4.846750369276219e-05,
      "loss": 0.1051,
      "step": 216
    },
    {
      "epoch": 0.08,
      "learning_rate": 4.84582717872969e-05,
      "loss": 0.0592,
      "step": 217
    },
    {
      "epoch": 0.08,
      "learning_rate": 4.844903988183161e-05,
      "loss": 0.0794,
      "step": 218
    },
    {
      "epoch": 0.08,
      "learning_rate": 4.843980797636632e-05,
      "loss": 0.0875,
      "step": 219
    },
    {
      "epoch": 0.08,
      "learning_rate": 4.843057607090103e-05,
      "loss": 0.0485,
      "step": 220
    },
    {
      "epoch": 0.08,
      "learning_rate": 4.842134416543575e-05,
      "loss": 0.1102,
      "step": 221
    },
    {
      "epoch": 0.08,
      "learning_rate": 4.841211225997046e-05,
      "loss": 0.1031,
      "step": 222
    },
    {
      "epoch": 0.08,
      "learning_rate": 4.840288035450517e-05,
      "loss": 0.5843,
      "step": 223
    },
    {
      "epoch": 0.08,
      "learning_rate": 4.839364844903989e-05,
      "loss": 0.2708,
      "step": 224
    },
    {
      "epoch": 0.08,
      "learning_rate": 4.838441654357459e-05,
      "loss": 0.2853,
      "step": 225
    },
    {
      "epoch": 0.08,
      "learning_rate": 4.83751846381093e-05,
      "loss": 0.233,
      "step": 226
    },
    {
      "epoch": 0.08,
      "learning_rate": 4.836595273264402e-05,
      "loss": 0.6123,
      "step": 227
    },
    {
      "epoch": 0.08,
      "learning_rate": 4.835672082717873e-05,
      "loss": 0.2169,
      "step": 228
    },
    {
      "epoch": 0.08,
      "learning_rate": 4.834748892171345e-05,
      "loss": 0.1015,
      "step": 229
    },
    {
      "epoch": 0.08,
      "learning_rate": 4.833825701624816e-05,
      "loss": 0.0832,
      "step": 230
    },
    {
      "epoch": 0.08,
      "learning_rate": 4.832902511078287e-05,
      "loss": 0.2653,
      "step": 231
    },
    {
      "epoch": 0.08,
      "learning_rate": 4.831979320531758e-05,
      "loss": 0.0655,
      "step": 232
    },
    {
      "epoch": 0.09,
      "learning_rate": 4.831056129985229e-05,
      "loss": 0.0982,
      "step": 233
    },
    {
      "epoch": 0.09,
      "learning_rate": 4.8301329394387e-05,
      "loss": 0.0432,
      "step": 234
    },
    {
      "epoch": 0.09,
      "learning_rate": 4.829209748892172e-05,
      "loss": 0.1143,
      "step": 235
    },
    {
      "epoch": 0.09,
      "learning_rate": 4.828286558345643e-05,
      "loss": 0.235,
      "step": 236
    },
    {
      "epoch": 0.09,
      "learning_rate": 4.827363367799114e-05,
      "loss": 0.1869,
      "step": 237
    },
    {
      "epoch": 0.09,
      "learning_rate": 4.826440177252585e-05,
      "loss": 0.0507,
      "step": 238
    },
    {
      "epoch": 0.09,
      "learning_rate": 4.825516986706056e-05,
      "loss": 0.1097,
      "step": 239
    },
    {
      "epoch": 0.09,
      "learning_rate": 4.824593796159528e-05,
      "loss": 0.036,
      "step": 240
    },
    {
      "epoch": 0.09,
      "learning_rate": 4.823670605612999e-05,
      "loss": 0.4278,
      "step": 241
    },
    {
      "epoch": 0.09,
      "learning_rate": 4.82274741506647e-05,
      "loss": 0.392,
      "step": 242
    },
    {
      "epoch": 0.09,
      "learning_rate": 4.8218242245199416e-05,
      "loss": 0.0121,
      "step": 243
    },
    {
      "epoch": 0.09,
      "learning_rate": 4.820901033973412e-05,
      "loss": 0.497,
      "step": 244
    },
    {
      "epoch": 0.09,
      "learning_rate": 4.819977843426883e-05,
      "loss": 0.0468,
      "step": 245
    },
    {
      "epoch": 0.09,
      "learning_rate": 4.819054652880355e-05,
      "loss": 0.4706,
      "step": 246
    },
    {
      "epoch": 0.09,
      "learning_rate": 4.818131462333826e-05,
      "loss": 0.1119,
      "step": 247
    },
    {
      "epoch": 0.09,
      "learning_rate": 4.8172082717872976e-05,
      "loss": 0.1022,
      "step": 248
    },
    {
      "epoch": 0.09,
      "learning_rate": 4.8162850812407686e-05,
      "loss": 0.2688,
      "step": 249
    },
    {
      "epoch": 0.09,
      "learning_rate": 4.81536189069424e-05,
      "loss": 0.0828,
      "step": 250
    },
    {
      "epoch": 0.09,
      "learning_rate": 4.814438700147711e-05,
      "loss": 0.1772,
      "step": 251
    },
    {
      "epoch": 0.09,
      "learning_rate": 4.813515509601182e-05,
      "loss": 0.3958,
      "step": 252
    },
    {
      "epoch": 0.09,
      "learning_rate": 4.812592319054653e-05,
      "loss": 0.7139,
      "step": 253
    },
    {
      "epoch": 0.09,
      "learning_rate": 4.8116691285081246e-05,
      "loss": 0.1215,
      "step": 254
    },
    {
      "epoch": 0.09,
      "learning_rate": 4.8107459379615956e-05,
      "loss": 0.2505,
      "step": 255
    },
    {
      "epoch": 0.09,
      "learning_rate": 4.809822747415067e-05,
      "loss": 0.2115,
      "step": 256
    },
    {
      "epoch": 0.09,
      "learning_rate": 4.808899556868538e-05,
      "loss": 0.1075,
      "step": 257
    },
    {
      "epoch": 0.09,
      "learning_rate": 4.807976366322009e-05,
      "loss": 0.1792,
      "step": 258
    },
    {
      "epoch": 0.09,
      "learning_rate": 4.8070531757754805e-05,
      "loss": 0.1301,
      "step": 259
    },
    {
      "epoch": 0.1,
      "learning_rate": 4.8061299852289516e-05,
      "loss": 0.2274,
      "step": 260
    },
    {
      "epoch": 0.1,
      "learning_rate": 4.8052067946824226e-05,
      "loss": 0.2079,
      "step": 261
    },
    {
      "epoch": 0.1,
      "learning_rate": 4.8042836041358944e-05,
      "loss": 0.565,
      "step": 262
    },
    {
      "epoch": 0.1,
      "learning_rate": 4.803360413589365e-05,
      "loss": 0.0785,
      "step": 263
    },
    {
      "epoch": 0.1,
      "learning_rate": 4.802437223042836e-05,
      "loss": 0.5263,
      "step": 264
    },
    {
      "epoch": 0.1,
      "learning_rate": 4.8015140324963075e-05,
      "loss": 0.1025,
      "step": 265
    },
    {
      "epoch": 0.1,
      "learning_rate": 4.8005908419497786e-05,
      "loss": 0.0875,
      "step": 266
    },
    {
      "epoch": 0.1,
      "learning_rate": 4.7996676514032496e-05,
      "loss": 0.2589,
      "step": 267
    },
    {
      "epoch": 0.1,
      "learning_rate": 4.7987444608567214e-05,
      "loss": 0.1881,
      "step": 268
    },
    {
      "epoch": 0.1,
      "learning_rate": 4.7978212703101924e-05,
      "loss": 0.0965,
      "step": 269
    },
    {
      "epoch": 0.1,
      "learning_rate": 4.7968980797636635e-05,
      "loss": 0.1144,
      "step": 270
    },
    {
      "epoch": 0.1,
      "learning_rate": 4.7959748892171345e-05,
      "loss": 0.1756,
      "step": 271
    },
    {
      "epoch": 0.1,
      "learning_rate": 4.7950516986706056e-05,
      "loss": 0.2536,
      "step": 272
    },
    {
      "epoch": 0.1,
      "learning_rate": 4.794128508124077e-05,
      "loss": 0.4324,
      "step": 273
    },
    {
      "epoch": 0.1,
      "learning_rate": 4.7932053175775484e-05,
      "loss": 0.2434,
      "step": 274
    },
    {
      "epoch": 0.1,
      "learning_rate": 4.7922821270310194e-05,
      "loss": 0.4039,
      "step": 275
    },
    {
      "epoch": 0.1,
      "learning_rate": 4.7913589364844905e-05,
      "loss": 0.0454,
      "step": 276
    },
    {
      "epoch": 0.1,
      "learning_rate": 4.7904357459379616e-05,
      "loss": 0.0479,
      "step": 277
    },
    {
      "epoch": 0.1,
      "learning_rate": 4.789512555391433e-05,
      "loss": 0.0569,
      "step": 278
    },
    {
      "epoch": 0.1,
      "learning_rate": 4.7885893648449043e-05,
      "loss": 0.4648,
      "step": 279
    },
    {
      "epoch": 0.1,
      "learning_rate": 4.7876661742983754e-05,
      "loss": 0.111,
      "step": 280
    },
    {
      "epoch": 0.1,
      "learning_rate": 4.786742983751847e-05,
      "loss": 0.2648,
      "step": 281
    },
    {
      "epoch": 0.1,
      "learning_rate": 4.7858197932053175e-05,
      "loss": 0.3365,
      "step": 282
    },
    {
      "epoch": 0.1,
      "learning_rate": 4.7848966026587886e-05,
      "loss": 0.1649,
      "step": 283
    },
    {
      "epoch": 0.1,
      "learning_rate": 4.78397341211226e-05,
      "loss": 0.1663,
      "step": 284
    },
    {
      "epoch": 0.1,
      "learning_rate": 4.7830502215657313e-05,
      "loss": 0.0741,
      "step": 285
    },
    {
      "epoch": 0.1,
      "learning_rate": 4.7821270310192024e-05,
      "loss": 0.3465,
      "step": 286
    },
    {
      "epoch": 0.11,
      "learning_rate": 4.781203840472674e-05,
      "loss": 0.0245,
      "step": 287
    },
    {
      "epoch": 0.11,
      "learning_rate": 4.780280649926145e-05,
      "loss": 0.1093,
      "step": 288
    },
    {
      "epoch": 0.11,
      "learning_rate": 4.779357459379616e-05,
      "loss": 0.0726,
      "step": 289
    },
    {
      "epoch": 0.11,
      "learning_rate": 4.778434268833087e-05,
      "loss": 0.2307,
      "step": 290
    },
    {
      "epoch": 0.11,
      "learning_rate": 4.7775110782865584e-05,
      "loss": 0.0388,
      "step": 291
    },
    {
      "epoch": 0.11,
      "learning_rate": 4.77658788774003e-05,
      "loss": 0.0088,
      "step": 292
    },
    {
      "epoch": 0.11,
      "learning_rate": 4.775664697193501e-05,
      "loss": 0.3141,
      "step": 293
    },
    {
      "epoch": 0.11,
      "learning_rate": 4.774741506646972e-05,
      "loss": 0.105,
      "step": 294
    },
    {
      "epoch": 0.11,
      "learning_rate": 4.773818316100443e-05,
      "loss": 0.3587,
      "step": 295
    },
    {
      "epoch": 0.11,
      "learning_rate": 4.772895125553914e-05,
      "loss": 0.0906,
      "step": 296
    },
    {
      "epoch": 0.11,
      "learning_rate": 4.7719719350073854e-05,
      "loss": 0.0985,
      "step": 297
    },
    {
      "epoch": 0.11,
      "learning_rate": 4.771048744460857e-05,
      "loss": 0.0761,
      "step": 298
    },
    {
      "epoch": 0.11,
      "learning_rate": 4.770125553914328e-05,
      "loss": 0.0228,
      "step": 299
    },
    {
      "epoch": 0.11,
      "learning_rate": 4.7692023633678e-05,
      "loss": 0.257,
      "step": 300
    },
    {
      "epoch": 0.11,
      "learning_rate": 4.76827917282127e-05,
      "loss": 0.5474,
      "step": 301
    },
    {
      "epoch": 0.11,
      "learning_rate": 4.767355982274741e-05,
      "loss": 0.2556,
      "step": 302
    },
    {
      "epoch": 0.11,
      "learning_rate": 4.766432791728213e-05,
      "loss": 0.4549,
      "step": 303
    },
    {
      "epoch": 0.11,
      "learning_rate": 4.765509601181684e-05,
      "loss": 0.3563,
      "step": 304
    },
    {
      "epoch": 0.11,
      "learning_rate": 4.764586410635155e-05,
      "loss": 0.1395,
      "step": 305
    },
    {
      "epoch": 0.11,
      "learning_rate": 4.763663220088627e-05,
      "loss": 0.192,
      "step": 306
    },
    {
      "epoch": 0.11,
      "learning_rate": 4.762740029542098e-05,
      "loss": 0.1645,
      "step": 307
    },
    {
      "epoch": 0.11,
      "learning_rate": 4.761816838995569e-05,
      "loss": 0.1052,
      "step": 308
    },
    {
      "epoch": 0.11,
      "learning_rate": 4.76089364844904e-05,
      "loss": 0.0597,
      "step": 309
    },
    {
      "epoch": 0.11,
      "learning_rate": 4.759970457902511e-05,
      "loss": 0.3124,
      "step": 310
    },
    {
      "epoch": 0.11,
      "learning_rate": 4.759047267355983e-05,
      "loss": 0.0139,
      "step": 311
    },
    {
      "epoch": 0.11,
      "learning_rate": 4.758124076809454e-05,
      "loss": 0.0083,
      "step": 312
    },
    {
      "epoch": 0.11,
      "learning_rate": 4.757200886262925e-05,
      "loss": 0.3064,
      "step": 313
    },
    {
      "epoch": 0.11,
      "learning_rate": 4.756277695716396e-05,
      "loss": 0.1306,
      "step": 314
    },
    {
      "epoch": 0.12,
      "learning_rate": 4.755354505169867e-05,
      "loss": 0.0392,
      "step": 315
    },
    {
      "epoch": 0.12,
      "learning_rate": 4.754431314623338e-05,
      "loss": 0.0951,
      "step": 316
    },
    {
      "epoch": 0.12,
      "learning_rate": 4.75350812407681e-05,
      "loss": 0.4908,
      "step": 317
    },
    {
      "epoch": 0.12,
      "learning_rate": 4.752584933530281e-05,
      "loss": 0.4677,
      "step": 318
    },
    {
      "epoch": 0.12,
      "learning_rate": 4.7516617429837526e-05,
      "loss": 0.3966,
      "step": 319
    },
    {
      "epoch": 0.12,
      "learning_rate": 4.750738552437223e-05,
      "loss": 0.0201,
      "step": 320
    },
    {
      "epoch": 0.12,
      "learning_rate": 4.749815361890694e-05,
      "loss": 0.2814,
      "step": 321
    },
    {
      "epoch": 0.12,
      "learning_rate": 4.748892171344166e-05,
      "loss": 0.0996,
      "step": 322
    },
    {
      "epoch": 0.12,
      "learning_rate": 4.747968980797637e-05,
      "loss": 0.0845,
      "step": 323
    },
    {
      "epoch": 0.12,
      "learning_rate": 4.747045790251108e-05,
      "loss": 0.0394,
      "step": 324
    },
    {
      "epoch": 0.12,
      "learning_rate": 4.7461225997045796e-05,
      "loss": 0.3497,
      "step": 325
    },
    {
      "epoch": 0.12,
      "learning_rate": 4.745199409158051e-05,
      "loss": 0.319,
      "step": 326
    },
    {
      "epoch": 0.12,
      "learning_rate": 4.744276218611521e-05,
      "loss": 0.6137,
      "step": 327
    },
    {
      "epoch": 0.12,
      "learning_rate": 4.743353028064993e-05,
      "loss": 0.0112,
      "step": 328
    },
    {
      "epoch": 0.12,
      "learning_rate": 4.742429837518464e-05,
      "loss": 0.0226,
      "step": 329
    },
    {
      "epoch": 0.12,
      "learning_rate": 4.7415066469719356e-05,
      "loss": 0.0148,
      "step": 330
    },
    {
      "epoch": 0.12,
      "learning_rate": 4.7405834564254066e-05,
      "loss": 0.5004,
      "step": 331
    },
    {
      "epoch": 0.12,
      "learning_rate": 4.739660265878878e-05,
      "loss": 0.2844,
      "step": 332
    },
    {
      "epoch": 0.12,
      "learning_rate": 4.738737075332349e-05,
      "loss": 0.0917,
      "step": 333
    },
    {
      "epoch": 0.12,
      "learning_rate": 4.73781388478582e-05,
      "loss": 0.2199,
      "step": 334
    },
    {
      "epoch": 0.12,
      "learning_rate": 4.736890694239291e-05,
      "loss": 0.0187,
      "step": 335
    },
    {
      "epoch": 0.12,
      "learning_rate": 4.7359675036927626e-05,
      "loss": 0.0059,
      "step": 336
    },
    {
      "epoch": 0.12,
      "learning_rate": 4.7350443131462337e-05,
      "loss": 0.707,
      "step": 337
    },
    {
      "epoch": 0.12,
      "learning_rate": 4.7341211225997054e-05,
      "loss": 0.4906,
      "step": 338
    },
    {
      "epoch": 0.12,
      "learning_rate": 4.733197932053176e-05,
      "loss": 0.0458,
      "step": 339
    },
    {
      "epoch": 0.12,
      "learning_rate": 4.732274741506647e-05,
      "loss": 0.8585,
      "step": 340
    },
    {
      "epoch": 0.12,
      "learning_rate": 4.7313515509601186e-05,
      "loss": 0.0559,
      "step": 341
    },
    {
      "epoch": 0.13,
      "learning_rate": 4.7304283604135896e-05,
      "loss": 0.4183,
      "step": 342
    },
    {
      "epoch": 0.13,
      "learning_rate": 4.729505169867061e-05,
      "loss": 0.0369,
      "step": 343
    },
    {
      "epoch": 0.13,
      "learning_rate": 4.7285819793205324e-05,
      "loss": 0.509,
      "step": 344
    },
    {
      "epoch": 0.13,
      "learning_rate": 4.7276587887740034e-05,
      "loss": 0.3659,
      "step": 345
    },
    {
      "epoch": 0.13,
      "learning_rate": 4.726735598227474e-05,
      "loss": 0.3091,
      "step": 346
    },
    {
      "epoch": 0.13,
      "learning_rate": 4.7258124076809456e-05,
      "loss": 0.2513,
      "step": 347
    },
    {
      "epoch": 0.13,
      "learning_rate": 4.7248892171344166e-05,
      "loss": 0.6178,
      "step": 348
    },
    {
      "epoch": 0.13,
      "learning_rate": 4.7239660265878883e-05,
      "loss": 0.3908,
      "step": 349
    },
    {
      "epoch": 0.13,
      "learning_rate": 4.7230428360413594e-05,
      "loss": 0.2756,
      "step": 350
    },
    {
      "epoch": 0.13,
      "learning_rate": 4.7221196454948305e-05,
      "loss": 0.1072,
      "step": 351
    },
    {
      "epoch": 0.13,
      "learning_rate": 4.7211964549483015e-05,
      "loss": 0.2133,
      "step": 352
    },
    {
      "epoch": 0.13,
      "learning_rate": 4.7202732644017726e-05,
      "loss": 0.2629,
      "step": 353
    },
    {
      "epoch": 0.13,
      "learning_rate": 4.7193500738552436e-05,
      "loss": 0.149,
      "step": 354
    },
    {
      "epoch": 0.13,
      "learning_rate": 4.7184268833087154e-05,
      "loss": 0.4268,
      "step": 355
    },
    {
      "epoch": 0.13,
      "learning_rate": 4.7175036927621864e-05,
      "loss": 0.2087,
      "step": 356
    },
    {
      "epoch": 0.13,
      "learning_rate": 4.7165805022156575e-05,
      "loss": 0.1505,
      "step": 357
    },
    {
      "epoch": 0.13,
      "learning_rate": 4.7156573116691285e-05,
      "loss": 0.0903,
      "step": 358
    },
    {
      "epoch": 0.13,
      "learning_rate": 4.7147341211225996e-05,
      "loss": 0.273,
      "step": 359
    },
    {
      "epoch": 0.13,
      "learning_rate": 4.713810930576071e-05,
      "loss": 0.1081,
      "step": 360
    },
    {
      "epoch": 0.13,
      "learning_rate": 4.7128877400295424e-05,
      "loss": 0.049,
      "step": 361
    },
    {
      "epoch": 0.13,
      "learning_rate": 4.7119645494830134e-05,
      "loss": 0.4231,
      "step": 362
    },
    {
      "epoch": 0.13,
      "learning_rate": 4.711041358936485e-05,
      "loss": 0.0588,
      "step": 363
    },
    {
      "epoch": 0.13,
      "learning_rate": 4.710118168389956e-05,
      "loss": 0.0249,
      "step": 364
    },
    {
      "epoch": 0.13,
      "learning_rate": 4.7091949778434266e-05,
      "loss": 0.1574,
      "step": 365
    },
    {
      "epoch": 0.13,
      "learning_rate": 4.708271787296898e-05,
      "loss": 0.062,
      "step": 366
    },
    {
      "epoch": 0.13,
      "learning_rate": 4.7073485967503694e-05,
      "loss": 0.3816,
      "step": 367
    },
    {
      "epoch": 0.13,
      "learning_rate": 4.7064254062038404e-05,
      "loss": 0.1578,
      "step": 368
    },
    {
      "epoch": 0.14,
      "learning_rate": 4.705502215657312e-05,
      "loss": 0.2193,
      "step": 369
    },
    {
      "epoch": 0.14,
      "learning_rate": 4.704579025110783e-05,
      "loss": 0.321,
      "step": 370
    },
    {
      "epoch": 0.14,
      "learning_rate": 4.703655834564254e-05,
      "loss": 0.031,
      "step": 371
    },
    {
      "epoch": 0.14,
      "learning_rate": 4.702732644017725e-05,
      "loss": 0.0674,
      "step": 372
    },
    {
      "epoch": 0.14,
      "learning_rate": 4.7018094534711964e-05,
      "loss": 0.0334,
      "step": 373
    },
    {
      "epoch": 0.14,
      "learning_rate": 4.700886262924668e-05,
      "loss": 0.5481,
      "step": 374
    },
    {
      "epoch": 0.14,
      "learning_rate": 4.699963072378139e-05,
      "loss": 0.5161,
      "step": 375
    },
    {
      "epoch": 0.14,
      "learning_rate": 4.69903988183161e-05,
      "loss": 0.2968,
      "step": 376
    },
    {
      "epoch": 0.14,
      "learning_rate": 4.698116691285081e-05,
      "loss": 0.0177,
      "step": 377
    },
    {
      "epoch": 0.14,
      "learning_rate": 4.697193500738552e-05,
      "loss": 0.4802,
      "step": 378
    },
    {
      "epoch": 0.14,
      "learning_rate": 4.696270310192024e-05,
      "loss": 0.0079,
      "step": 379
    },
    {
      "epoch": 0.14,
      "learning_rate": 4.695347119645495e-05,
      "loss": 0.2142,
      "step": 380
    },
    {
      "epoch": 0.14,
      "learning_rate": 4.694423929098966e-05,
      "loss": 0.1825,
      "step": 381
    },
    {
      "epoch": 0.14,
      "learning_rate": 4.693500738552438e-05,
      "loss": 0.0116,
      "step": 382
    },
    {
      "epoch": 0.14,
      "learning_rate": 4.692577548005909e-05,
      "loss": 0.0125,
      "step": 383
    },
    {
      "epoch": 0.14,
      "learning_rate": 4.691654357459379e-05,
      "loss": 0.4642,
      "step": 384
    },
    {
      "epoch": 0.14,
      "learning_rate": 4.690731166912851e-05,
      "loss": 0.3389,
      "step": 385
    },
    {
      "epoch": 0.14,
      "learning_rate": 4.689807976366322e-05,
      "loss": 0.0094,
      "step": 386
    },
    {
      "epoch": 0.14,
      "learning_rate": 4.688884785819793e-05,
      "loss": 0.3949,
      "step": 387
    },
    {
      "epoch": 0.14,
      "learning_rate": 4.687961595273265e-05,
      "loss": 0.8848,
      "step": 388
    },
    {
      "epoch": 0.14,
      "learning_rate": 4.687038404726736e-05,
      "loss": 0.3533,
      "step": 389
    },
    {
      "epoch": 0.14,
      "learning_rate": 4.686115214180207e-05,
      "loss": 0.3494,
      "step": 390
    },
    {
      "epoch": 0.14,
      "learning_rate": 4.685192023633678e-05,
      "loss": 0.6242,
      "step": 391
    },
    {
      "epoch": 0.14,
      "learning_rate": 4.684268833087149e-05,
      "loss": 0.3346,
      "step": 392
    },
    {
      "epoch": 0.14,
      "learning_rate": 4.683345642540621e-05,
      "loss": 0.1879,
      "step": 393
    },
    {
      "epoch": 0.14,
      "learning_rate": 4.682422451994092e-05,
      "loss": 0.189,
      "step": 394
    },
    {
      "epoch": 0.14,
      "learning_rate": 4.681499261447563e-05,
      "loss": 0.018,
      "step": 395
    },
    {
      "epoch": 0.14,
      "learning_rate": 4.680576070901034e-05,
      "loss": 0.2368,
      "step": 396
    },
    {
      "epoch": 0.15,
      "learning_rate": 4.679652880354505e-05,
      "loss": 0.1815,
      "step": 397
    },
    {
      "epoch": 0.15,
      "learning_rate": 4.678729689807976e-05,
      "loss": 0.095,
      "step": 398
    },
    {
      "epoch": 0.15,
      "learning_rate": 4.677806499261448e-05,
      "loss": 0.1989,
      "step": 399
    },
    {
      "epoch": 0.15,
      "learning_rate": 4.676883308714919e-05,
      "loss": 0.1428,
      "step": 400
    },
    {
      "epoch": 0.15,
      "learning_rate": 4.6759601181683907e-05,
      "loss": 0.2218,
      "step": 401
    },
    {
      "epoch": 0.15,
      "learning_rate": 4.675036927621862e-05,
      "loss": 0.1987,
      "step": 402
    },
    {
      "epoch": 0.15,
      "learning_rate": 4.674113737075332e-05,
      "loss": 0.1087,
      "step": 403
    },
    {
      "epoch": 0.15,
      "learning_rate": 4.673190546528804e-05,
      "loss": 0.0508,
      "step": 404
    },
    {
      "epoch": 0.15,
      "learning_rate": 4.672267355982275e-05,
      "loss": 0.0605,
      "step": 405
    },
    {
      "epoch": 0.15,
      "learning_rate": 4.671344165435746e-05,
      "loss": 0.0498,
      "step": 406
    },
    {
      "epoch": 0.15,
      "learning_rate": 4.670420974889218e-05,
      "loss": 0.033,
      "step": 407
    },
    {
      "epoch": 0.15,
      "learning_rate": 4.669497784342689e-05,
      "loss": 0.1795,
      "step": 408
    },
    {
      "epoch": 0.15,
      "learning_rate": 4.66857459379616e-05,
      "loss": 0.2933,
      "step": 409
    },
    {
      "epoch": 0.15,
      "learning_rate": 4.667651403249631e-05,
      "loss": 0.2665,
      "step": 410
    },
    {
      "epoch": 0.15,
      "learning_rate": 4.666728212703102e-05,
      "loss": 0.0766,
      "step": 411
    },
    {
      "epoch": 0.15,
      "learning_rate": 4.6658050221565736e-05,
      "loss": 0.5206,
      "step": 412
    },
    {
      "epoch": 0.15,
      "learning_rate": 4.664881831610045e-05,
      "loss": 0.0343,
      "step": 413
    },
    {
      "epoch": 0.15,
      "learning_rate": 4.663958641063516e-05,
      "loss": 0.0282,
      "step": 414
    },
    {
      "epoch": 0.15,
      "learning_rate": 4.663035450516987e-05,
      "loss": 0.2619,
      "step": 415
    },
    {
      "epoch": 0.15,
      "learning_rate": 4.662112259970458e-05,
      "loss": 0.5237,
      "step": 416
    },
    {
      "epoch": 0.15,
      "learning_rate": 4.661189069423929e-05,
      "loss": 0.1532,
      "step": 417
    },
    {
      "epoch": 0.15,
      "learning_rate": 4.6602658788774006e-05,
      "loss": 0.0921,
      "step": 418
    },
    {
      "epoch": 0.15,
      "learning_rate": 4.659342688330872e-05,
      "loss": 0.0894,
      "step": 419
    },
    {
      "epoch": 0.15,
      "learning_rate": 4.6584194977843434e-05,
      "loss": 0.1797,
      "step": 420
    },
    {
      "epoch": 0.15,
      "learning_rate": 4.6574963072378145e-05,
      "loss": 0.2062,
      "step": 421
    },
    {
      "epoch": 0.15,
      "learning_rate": 4.656573116691285e-05,
      "loss": 0.1825,
      "step": 422
    },
    {
      "epoch": 0.15,
      "learning_rate": 4.6556499261447566e-05,
      "loss": 0.0837,
      "step": 423
    },
    {
      "epoch": 0.16,
      "learning_rate": 4.6547267355982276e-05,
      "loss": 0.028,
      "step": 424
    },
    {
      "epoch": 0.16,
      "learning_rate": 4.653803545051699e-05,
      "loss": 0.0367,
      "step": 425
    },
    {
      "epoch": 0.16,
      "learning_rate": 4.6528803545051704e-05,
      "loss": 0.2524,
      "step": 426
    },
    {
      "epoch": 0.16,
      "learning_rate": 4.6519571639586415e-05,
      "loss": 0.0211,
      "step": 427
    },
    {
      "epoch": 0.16,
      "learning_rate": 4.6510339734121125e-05,
      "loss": 0.1769,
      "step": 428
    },
    {
      "epoch": 0.16,
      "learning_rate": 4.6501107828655836e-05,
      "loss": 0.1995,
      "step": 429
    },
    {
      "epoch": 0.16,
      "learning_rate": 4.6491875923190546e-05,
      "loss": 0.0135,
      "step": 430
    },
    {
      "epoch": 0.16,
      "learning_rate": 4.6482644017725264e-05,
      "loss": 0.0304,
      "step": 431
    },
    {
      "epoch": 0.16,
      "learning_rate": 4.6473412112259974e-05,
      "loss": 0.2289,
      "step": 432
    },
    {
      "epoch": 0.16,
      "learning_rate": 4.6464180206794685e-05,
      "loss": 0.4831,
      "step": 433
    },
    {
      "epoch": 0.16,
      "learning_rate": 4.6454948301329395e-05,
      "loss": 0.3024,
      "step": 434
    },
    {
      "epoch": 0.16,
      "learning_rate": 4.6445716395864106e-05,
      "loss": 0.3362,
      "step": 435
    },
    {
      "epoch": 0.16,
      "learning_rate": 4.6436484490398816e-05,
      "loss": 0.5694,
      "step": 436
    },
    {
      "epoch": 0.16,
      "learning_rate": 4.6427252584933534e-05,
      "loss": 0.2486,
      "step": 437
    },
    {
      "epoch": 0.16,
      "learning_rate": 4.6418020679468244e-05,
      "loss": 0.0289,
      "step": 438
    },
    {
      "epoch": 0.16,
      "learning_rate": 4.6408788774002955e-05,
      "loss": 0.2093,
      "step": 439
    },
    {
      "epoch": 0.16,
      "learning_rate": 4.639955686853767e-05,
      "loss": 0.3298,
      "step": 440
    },
    {
      "epoch": 0.16,
      "learning_rate": 4.6390324963072376e-05,
      "loss": 0.4126,
      "step": 441
    },
    {
      "epoch": 0.16,
      "learning_rate": 4.638109305760709e-05,
      "loss": 0.1704,
      "step": 442
    },
    {
      "epoch": 0.16,
      "learning_rate": 4.6371861152141804e-05,
      "loss": 0.0583,
      "step": 443
    },
    {
      "epoch": 0.16,
      "learning_rate": 4.6362629246676514e-05,
      "loss": 0.0736,
      "step": 444
    },
    {
      "epoch": 0.16,
      "learning_rate": 4.635339734121123e-05,
      "loss": 0.644,
      "step": 445
    },
    {
      "epoch": 0.16,
      "learning_rate": 4.634416543574594e-05,
      "loss": 0.4253,
      "step": 446
    },
    {
      "epoch": 0.16,
      "learning_rate": 4.633493353028065e-05,
      "loss": 0.2098,
      "step": 447
    },
    {
      "epoch": 0.16,
      "learning_rate": 4.632570162481536e-05,
      "loss": 0.3045,
      "step": 448
    },
    {
      "epoch": 0.16,
      "learning_rate": 4.6316469719350074e-05,
      "loss": 0.0739,
      "step": 449
    },
    {
      "epoch": 0.16,
      "learning_rate": 4.630723781388479e-05,
      "loss": 0.1519,
      "step": 450
    },
    {
      "epoch": 0.17,
      "learning_rate": 4.62980059084195e-05,
      "loss": 0.1034,
      "step": 451
    },
    {
      "epoch": 0.17,
      "learning_rate": 4.628877400295421e-05,
      "loss": 0.0683,
      "step": 452
    },
    {
      "epoch": 0.17,
      "learning_rate": 4.627954209748892e-05,
      "loss": 0.1755,
      "step": 453
    },
    {
      "epoch": 0.17,
      "learning_rate": 4.6270310192023633e-05,
      "loss": 0.0385,
      "step": 454
    },
    {
      "epoch": 0.17,
      "learning_rate": 4.6261078286558344e-05,
      "loss": 0.3912,
      "step": 455
    },
    {
      "epoch": 0.17,
      "learning_rate": 4.625184638109306e-05,
      "loss": 0.0197,
      "step": 456
    },
    {
      "epoch": 0.17,
      "learning_rate": 4.624261447562777e-05,
      "loss": 0.3052,
      "step": 457
    },
    {
      "epoch": 0.17,
      "learning_rate": 4.623338257016248e-05,
      "loss": 0.0146,
      "step": 458
    },
    {
      "epoch": 0.17,
      "learning_rate": 4.62241506646972e-05,
      "loss": 0.0221,
      "step": 459
    },
    {
      "epoch": 0.17,
      "learning_rate": 4.6214918759231903e-05,
      "loss": 0.0175,
      "step": 460
    },
    {
      "epoch": 0.17,
      "learning_rate": 4.620568685376662e-05,
      "loss": 0.2946,
      "step": 461
    },
    {
      "epoch": 0.17,
      "learning_rate": 4.619645494830133e-05,
      "loss": 0.0272,
      "step": 462
    },
    {
      "epoch": 0.17,
      "learning_rate": 4.618722304283604e-05,
      "loss": 0.2952,
      "step": 463
    },
    {
      "epoch": 0.17,
      "learning_rate": 4.617799113737076e-05,
      "loss": 0.1616,
      "step": 464
    },
    {
      "epoch": 0.17,
      "learning_rate": 4.616875923190547e-05,
      "loss": 0.802,
      "step": 465
    },
    {
      "epoch": 0.17,
      "learning_rate": 4.615952732644018e-05,
      "loss": 0.0427,
      "step": 466
    },
    {
      "epoch": 0.17,
      "learning_rate": 4.615029542097489e-05,
      "loss": 0.0115,
      "step": 467
    },
    {
      "epoch": 0.17,
      "learning_rate": 4.61410635155096e-05,
      "loss": 0.4555,
      "step": 468
    },
    {
      "epoch": 0.17,
      "learning_rate": 4.613183161004431e-05,
      "loss": 0.3903,
      "step": 469
    },
    {
      "epoch": 0.17,
      "learning_rate": 4.612259970457903e-05,
      "loss": 0.0095,
      "step": 470
    },
    {
      "epoch": 0.17,
      "learning_rate": 4.611336779911374e-05,
      "loss": 0.006,
      "step": 471
    },
    {
      "epoch": 0.17,
      "learning_rate": 4.610413589364845e-05,
      "loss": 0.8774,
      "step": 472
    },
    {
      "epoch": 0.17,
      "learning_rate": 4.609490398818316e-05,
      "loss": 0.0341,
      "step": 473
    },
    {
      "epoch": 0.17,
      "learning_rate": 4.608567208271787e-05,
      "loss": 0.0118,
      "step": 474
    },
    {
      "epoch": 0.17,
      "learning_rate": 4.607644017725259e-05,
      "loss": 0.1543,
      "step": 475
    },
    {
      "epoch": 0.17,
      "learning_rate": 4.60672082717873e-05,
      "loss": 0.228,
      "step": 476
    },
    {
      "epoch": 0.17,
      "learning_rate": 4.605797636632201e-05,
      "loss": 0.0074,
      "step": 477
    },
    {
      "epoch": 0.17,
      "learning_rate": 4.604874446085673e-05,
      "loss": 0.0271,
      "step": 478
    },
    {
      "epoch": 0.18,
      "learning_rate": 4.603951255539143e-05,
      "loss": 0.2029,
      "step": 479
    },
    {
      "epoch": 0.18,
      "learning_rate": 4.603028064992615e-05,
      "loss": 0.0464,
      "step": 480
    },
    {
      "epoch": 0.18,
      "learning_rate": 4.602104874446086e-05,
      "loss": 0.0078,
      "step": 481
    },
    {
      "epoch": 0.18,
      "learning_rate": 4.601181683899557e-05,
      "loss": 0.1872,
      "step": 482
    },
    {
      "epoch": 0.18,
      "learning_rate": 4.600258493353029e-05,
      "loss": 0.1201,
      "step": 483
    },
    {
      "epoch": 0.18,
      "learning_rate": 4.5993353028065e-05,
      "loss": 0.3002,
      "step": 484
    },
    {
      "epoch": 0.18,
      "learning_rate": 4.598412112259971e-05,
      "loss": 0.0975,
      "step": 485
    },
    {
      "epoch": 0.18,
      "learning_rate": 4.597488921713442e-05,
      "loss": 0.0302,
      "step": 486
    },
    {
      "epoch": 0.18,
      "learning_rate": 4.596565731166913e-05,
      "loss": 0.1738,
      "step": 487
    },
    {
      "epoch": 0.18,
      "learning_rate": 4.595642540620384e-05,
      "loss": 0.0301,
      "step": 488
    },
    {
      "epoch": 0.18,
      "learning_rate": 4.594719350073856e-05,
      "loss": 0.2678,
      "step": 489
    },
    {
      "epoch": 0.18,
      "learning_rate": 4.593796159527327e-05,
      "loss": 0.3014,
      "step": 490
    },
    {
      "epoch": 0.18,
      "learning_rate": 4.592872968980798e-05,
      "loss": 0.2148,
      "step": 491
    },
    {
      "epoch": 0.18,
      "learning_rate": 4.591949778434269e-05,
      "loss": 0.0139,
      "step": 492
    },
    {
      "epoch": 0.18,
      "learning_rate": 4.59102658788774e-05,
      "loss": 0.0787,
      "step": 493
    },
    {
      "epoch": 0.18,
      "learning_rate": 4.5901033973412116e-05,
      "loss": 0.0091,
      "step": 494
    },
    {
      "epoch": 0.18,
      "learning_rate": 4.589180206794683e-05,
      "loss": 0.1429,
      "step": 495
    },
    {
      "epoch": 0.18,
      "learning_rate": 4.588257016248154e-05,
      "loss": 0.0109,
      "step": 496
    },
    {
      "epoch": 0.18,
      "learning_rate": 4.5873338257016255e-05,
      "loss": 0.0279,
      "step": 497
    },
    {
      "epoch": 0.18,
      "learning_rate": 4.586410635155096e-05,
      "loss": 0.5891,
      "step": 498
    },
    {
      "epoch": 0.18,
      "learning_rate": 4.585487444608567e-05,
      "loss": 0.0971,
      "step": 499
    },
    {
      "epoch": 0.18,
      "learning_rate": 4.5845642540620386e-05,
      "loss": 0.135,
      "step": 500
    },
    {
      "epoch": 0.18,
      "learning_rate": 4.58364106351551e-05,
      "loss": 0.0645,
      "step": 501
    },
    {
      "epoch": 0.18,
      "learning_rate": 4.5827178729689814e-05,
      "loss": 0.0145,
      "step": 502
    },
    {
      "epoch": 0.18,
      "learning_rate": 4.5817946824224525e-05,
      "loss": 0.3822,
      "step": 503
    },
    {
      "epoch": 0.18,
      "learning_rate": 4.5808714918759235e-05,
      "loss": 0.3499,
      "step": 504
    },
    {
      "epoch": 0.18,
      "learning_rate": 4.5799483013293946e-05,
      "loss": 0.399,
      "step": 505
    },
    {
      "epoch": 0.19,
      "learning_rate": 4.5790251107828657e-05,
      "loss": 0.0515,
      "step": 506
    },
    {
      "epoch": 0.19,
      "learning_rate": 4.578101920236337e-05,
      "loss": 0.1073,
      "step": 507
    },
    {
      "epoch": 0.19,
      "learning_rate": 4.5771787296898084e-05,
      "loss": 0.238,
      "step": 508
    },
    {
      "epoch": 0.19,
      "learning_rate": 4.5762555391432795e-05,
      "loss": 0.2476,
      "step": 509
    },
    {
      "epoch": 0.19,
      "learning_rate": 4.5753323485967505e-05,
      "loss": 0.0153,
      "step": 510
    },
    {
      "epoch": 0.19,
      "learning_rate": 4.5744091580502216e-05,
      "loss": 0.0769,
      "step": 511
    },
    {
      "epoch": 0.19,
      "learning_rate": 4.5734859675036927e-05,
      "loss": 0.0169,
      "step": 512
    },
    {
      "epoch": 0.19,
      "learning_rate": 4.5725627769571644e-05,
      "loss": 0.0197,
      "step": 513
    },
    {
      "epoch": 0.19,
      "learning_rate": 4.5716395864106354e-05,
      "loss": 0.0135,
      "step": 514
    },
    {
      "epoch": 0.19,
      "learning_rate": 4.5707163958641065e-05,
      "loss": 0.014,
      "step": 515
    },
    {
      "epoch": 0.19,
      "learning_rate": 4.569793205317578e-05,
      "loss": 0.2829,
      "step": 516
    },
    {
      "epoch": 0.19,
      "learning_rate": 4.5688700147710486e-05,
      "loss": 0.0091,
      "step": 517
    },
    {
      "epoch": 0.19,
      "learning_rate": 4.56794682422452e-05,
      "loss": 0.2166,
      "step": 518
    },
    {
      "epoch": 0.19,
      "learning_rate": 4.5670236336779914e-05,
      "loss": 0.1409,
      "step": 519
    },
    {
      "epoch": 0.19,
      "learning_rate": 4.5661004431314625e-05,
      "loss": 0.4492,
      "step": 520
    },
    {
      "epoch": 0.19,
      "learning_rate": 4.565177252584934e-05,
      "loss": 0.6128,
      "step": 521
    },
    {
      "epoch": 0.19,
      "learning_rate": 4.564254062038405e-05,
      "loss": 0.1108,
      "step": 522
    },
    {
      "epoch": 0.19,
      "learning_rate": 4.563330871491876e-05,
      "loss": 0.1372,
      "step": 523
    },
    {
      "epoch": 0.19,
      "learning_rate": 4.5624076809453473e-05,
      "loss": 0.2389,
      "step": 524
    },
    {
      "epoch": 0.19,
      "learning_rate": 4.5614844903988184e-05,
      "loss": 0.089,
      "step": 525
    },
    {
      "epoch": 0.19,
      "learning_rate": 4.5605612998522895e-05,
      "loss": 0.1443,
      "step": 526
    },
    {
      "epoch": 0.19,
      "learning_rate": 4.559638109305761e-05,
      "loss": 0.0096,
      "step": 527
    },
    {
      "epoch": 0.19,
      "learning_rate": 4.558714918759232e-05,
      "loss": 0.0787,
      "step": 528
    },
    {
      "epoch": 0.19,
      "learning_rate": 4.557791728212703e-05,
      "loss": 0.0493,
      "step": 529
    },
    {
      "epoch": 0.19,
      "learning_rate": 4.5568685376661744e-05,
      "loss": 0.3308,
      "step": 530
    },
    {
      "epoch": 0.19,
      "learning_rate": 4.5559453471196454e-05,
      "loss": 0.2417,
      "step": 531
    },
    {
      "epoch": 0.19,
      "learning_rate": 4.555022156573117e-05,
      "loss": 0.2008,
      "step": 532
    },
    {
      "epoch": 0.2,
      "learning_rate": 4.554098966026588e-05,
      "loss": 0.1094,
      "step": 533
    },
    {
      "epoch": 0.2,
      "learning_rate": 4.553175775480059e-05,
      "loss": 0.5293,
      "step": 534
    },
    {
      "epoch": 0.2,
      "learning_rate": 4.552252584933531e-05,
      "loss": 0.7033,
      "step": 535
    },
    {
      "epoch": 0.2,
      "learning_rate": 4.5513293943870014e-05,
      "loss": 0.1049,
      "step": 536
    },
    {
      "epoch": 0.2,
      "learning_rate": 4.5504062038404724e-05,
      "loss": 0.4175,
      "step": 537
    },
    {
      "epoch": 0.2,
      "learning_rate": 4.549483013293944e-05,
      "loss": 0.0879,
      "step": 538
    },
    {
      "epoch": 0.2,
      "learning_rate": 4.548559822747415e-05,
      "loss": 0.0293,
      "step": 539
    },
    {
      "epoch": 0.2,
      "learning_rate": 4.547636632200886e-05,
      "loss": 0.0672,
      "step": 540
    },
    {
      "epoch": 0.2,
      "learning_rate": 4.546713441654358e-05,
      "loss": 0.0361,
      "step": 541
    },
    {
      "epoch": 0.2,
      "learning_rate": 4.545790251107829e-05,
      "loss": 0.4993,
      "step": 542
    },
    {
      "epoch": 0.2,
      "learning_rate": 4.5448670605613e-05,
      "loss": 0.0139,
      "step": 543
    },
    {
      "epoch": 0.2,
      "learning_rate": 4.543943870014771e-05,
      "loss": 0.1334,
      "step": 544
    },
    {
      "epoch": 0.2,
      "learning_rate": 4.543020679468242e-05,
      "loss": 0.0159,
      "step": 545
    },
    {
      "epoch": 0.2,
      "learning_rate": 4.542097488921714e-05,
      "loss": 0.5574,
      "step": 546
    },
    {
      "epoch": 0.2,
      "learning_rate": 4.541174298375185e-05,
      "loss": 0.2237,
      "step": 547
    },
    {
      "epoch": 0.2,
      "learning_rate": 4.540251107828656e-05,
      "loss": 0.0671,
      "step": 548
    },
    {
      "epoch": 0.2,
      "learning_rate": 4.539327917282127e-05,
      "loss": 0.1734,
      "step": 549
    },
    {
      "epoch": 0.2,
      "learning_rate": 4.538404726735598e-05,
      "loss": 0.2198,
      "step": 550
    },
    {
      "epoch": 0.2,
      "learning_rate": 4.53748153618907e-05,
      "loss": 0.0285,
      "step": 551
    },
    {
      "epoch": 0.2,
      "learning_rate": 4.536558345642541e-05,
      "loss": 0.0131,
      "step": 552
    },
    {
      "epoch": 0.2,
      "learning_rate": 4.535635155096012e-05,
      "loss": 0.044,
      "step": 553
    },
    {
      "epoch": 0.2,
      "learning_rate": 4.534711964549484e-05,
      "loss": 0.057,
      "step": 554
    },
    {
      "epoch": 0.2,
      "learning_rate": 4.533788774002954e-05,
      "loss": 0.0346,
      "step": 555
    },
    {
      "epoch": 0.2,
      "learning_rate": 4.532865583456425e-05,
      "loss": 0.244,
      "step": 556
    },
    {
      "epoch": 0.2,
      "learning_rate": 4.531942392909897e-05,
      "loss": 0.0343,
      "step": 557
    },
    {
      "epoch": 0.2,
      "learning_rate": 4.531019202363368e-05,
      "loss": 0.3288,
      "step": 558
    },
    {
      "epoch": 0.2,
      "learning_rate": 4.530096011816839e-05,
      "loss": 0.5087,
      "step": 559
    },
    {
      "epoch": 0.2,
      "learning_rate": 4.529172821270311e-05,
      "loss": 0.0165,
      "step": 560
    },
    {
      "epoch": 0.21,
      "learning_rate": 4.528249630723782e-05,
      "loss": 0.2518,
      "step": 561
    },
    {
      "epoch": 0.21,
      "learning_rate": 4.527326440177253e-05,
      "loss": 0.1931,
      "step": 562
    },
    {
      "epoch": 0.21,
      "learning_rate": 4.526403249630724e-05,
      "loss": 0.0758,
      "step": 563
    },
    {
      "epoch": 0.21,
      "learning_rate": 4.525480059084195e-05,
      "loss": 0.1509,
      "step": 564
    },
    {
      "epoch": 0.21,
      "learning_rate": 4.524556868537667e-05,
      "loss": 0.3075,
      "step": 565
    },
    {
      "epoch": 0.21,
      "learning_rate": 4.523633677991138e-05,
      "loss": 0.0202,
      "step": 566
    },
    {
      "epoch": 0.21,
      "learning_rate": 4.522710487444609e-05,
      "loss": 0.1399,
      "step": 567
    },
    {
      "epoch": 0.21,
      "learning_rate": 4.52178729689808e-05,
      "loss": 0.3025,
      "step": 568
    },
    {
      "epoch": 0.21,
      "learning_rate": 4.520864106351551e-05,
      "loss": 0.0424,
      "step": 569
    },
    {
      "epoch": 0.21,
      "learning_rate": 4.519940915805022e-05,
      "loss": 0.0676,
      "step": 570
    },
    {
      "epoch": 0.21,
      "learning_rate": 4.519017725258494e-05,
      "loss": 0.1045,
      "step": 571
    },
    {
      "epoch": 0.21,
      "learning_rate": 4.518094534711965e-05,
      "loss": 0.1803,
      "step": 572
    },
    {
      "epoch": 0.21,
      "learning_rate": 4.5171713441654365e-05,
      "loss": 0.2651,
      "step": 573
    },
    {
      "epoch": 0.21,
      "learning_rate": 4.516248153618907e-05,
      "loss": 0.1026,
      "step": 574
    },
    {
      "epoch": 0.21,
      "learning_rate": 4.515324963072378e-05,
      "loss": 0.4029,
      "step": 575
    },
    {
      "epoch": 0.21,
      "learning_rate": 4.5144017725258497e-05,
      "loss": 0.2847,
      "step": 576
    },
    {
      "epoch": 0.21,
      "learning_rate": 4.513478581979321e-05,
      "loss": 0.3212,
      "step": 577
    },
    {
      "epoch": 0.21,
      "learning_rate": 4.512555391432792e-05,
      "loss": 0.4247,
      "step": 578
    },
    {
      "epoch": 0.21,
      "learning_rate": 4.5116322008862635e-05,
      "loss": 0.2085,
      "step": 579
    },
    {
      "epoch": 0.21,
      "learning_rate": 4.5107090103397346e-05,
      "loss": 0.1462,
      "step": 580
    },
    {
      "epoch": 0.21,
      "learning_rate": 4.5097858197932056e-05,
      "loss": 0.1609,
      "step": 581
    },
    {
      "epoch": 0.21,
      "learning_rate": 4.508862629246677e-05,
      "loss": 0.1025,
      "step": 582
    },
    {
      "epoch": 0.21,
      "learning_rate": 4.507939438700148e-05,
      "loss": 0.3727,
      "step": 583
    },
    {
      "epoch": 0.21,
      "learning_rate": 4.5070162481536195e-05,
      "loss": 0.3049,
      "step": 584
    },
    {
      "epoch": 0.21,
      "learning_rate": 4.5060930576070905e-05,
      "loss": 0.3419,
      "step": 585
    },
    {
      "epoch": 0.21,
      "learning_rate": 4.5051698670605616e-05,
      "loss": 0.4197,
      "step": 586
    },
    {
      "epoch": 0.21,
      "learning_rate": 4.5042466765140326e-05,
      "loss": 0.2907,
      "step": 587
    },
    {
      "epoch": 0.22,
      "learning_rate": 4.503323485967504e-05,
      "loss": 0.2249,
      "step": 588
    },
    {
      "epoch": 0.22,
      "learning_rate": 4.502400295420975e-05,
      "loss": 0.379,
      "step": 589
    },
    {
      "epoch": 0.22,
      "learning_rate": 4.5014771048744465e-05,
      "loss": 0.3288,
      "step": 590
    },
    {
      "epoch": 0.22,
      "learning_rate": 4.5005539143279175e-05,
      "loss": 0.0516,
      "step": 591
    },
    {
      "epoch": 0.22,
      "learning_rate": 4.499630723781389e-05,
      "loss": 0.0444,
      "step": 592
    },
    {
      "epoch": 0.22,
      "learning_rate": 4.4987075332348596e-05,
      "loss": 0.0606,
      "step": 593
    },
    {
      "epoch": 0.22,
      "learning_rate": 4.497784342688331e-05,
      "loss": 0.121,
      "step": 594
    },
    {
      "epoch": 0.22,
      "learning_rate": 4.4968611521418024e-05,
      "loss": 0.2166,
      "step": 595
    },
    {
      "epoch": 0.22,
      "learning_rate": 4.4959379615952735e-05,
      "loss": 0.0473,
      "step": 596
    },
    {
      "epoch": 0.22,
      "learning_rate": 4.4950147710487445e-05,
      "loss": 0.0162,
      "step": 597
    },
    {
      "epoch": 0.22,
      "learning_rate": 4.494091580502216e-05,
      "loss": 0.0615,
      "step": 598
    },
    {
      "epoch": 0.22,
      "learning_rate": 4.493168389955687e-05,
      "loss": 0.176,
      "step": 599
    },
    {
      "epoch": 0.22,
      "learning_rate": 4.492245199409158e-05,
      "loss": 0.0691,
      "step": 600
    },
    {
      "epoch": 0.22,
      "learning_rate": 4.4913220088626294e-05,
      "loss": 0.0724,
      "step": 601
    },
    {
      "epoch": 0.22,
      "learning_rate": 4.4903988183161005e-05,
      "loss": 0.1032,
      "step": 602
    },
    {
      "epoch": 0.22,
      "learning_rate": 4.489475627769572e-05,
      "loss": 0.0274,
      "step": 603
    },
    {
      "epoch": 0.22,
      "learning_rate": 4.488552437223043e-05,
      "loss": 0.0289,
      "step": 604
    },
    {
      "epoch": 0.22,
      "learning_rate": 4.487629246676514e-05,
      "loss": 0.1365,
      "step": 605
    },
    {
      "epoch": 0.22,
      "learning_rate": 4.4867060561299854e-05,
      "loss": 0.0134,
      "step": 606
    },
    {
      "epoch": 0.22,
      "learning_rate": 4.4857828655834564e-05,
      "loss": 0.204,
      "step": 607
    },
    {
      "epoch": 0.22,
      "learning_rate": 4.4848596750369275e-05,
      "loss": 0.3243,
      "step": 608
    },
    {
      "epoch": 0.22,
      "learning_rate": 4.483936484490399e-05,
      "loss": 0.0385,
      "step": 609
    },
    {
      "epoch": 0.22,
      "learning_rate": 4.48301329394387e-05,
      "loss": 0.258,
      "step": 610
    },
    {
      "epoch": 0.22,
      "learning_rate": 4.482090103397342e-05,
      "loss": 0.0069,
      "step": 611
    },
    {
      "epoch": 0.22,
      "learning_rate": 4.4811669128508124e-05,
      "loss": 0.1864,
      "step": 612
    },
    {
      "epoch": 0.22,
      "learning_rate": 4.4802437223042834e-05,
      "loss": 0.0067,
      "step": 613
    },
    {
      "epoch": 0.22,
      "learning_rate": 4.479320531757755e-05,
      "loss": 0.1177,
      "step": 614
    },
    {
      "epoch": 0.23,
      "learning_rate": 4.478397341211226e-05,
      "loss": 0.1708,
      "step": 615
    },
    {
      "epoch": 0.23,
      "learning_rate": 4.477474150664697e-05,
      "loss": 0.8025,
      "step": 616
    },
    {
      "epoch": 0.23,
      "learning_rate": 4.476550960118169e-05,
      "loss": 0.0238,
      "step": 617
    },
    {
      "epoch": 0.23,
      "learning_rate": 4.47562776957164e-05,
      "loss": 0.0137,
      "step": 618
    },
    {
      "epoch": 0.23,
      "learning_rate": 4.4747045790251104e-05,
      "loss": 0.2056,
      "step": 619
    },
    {
      "epoch": 0.23,
      "learning_rate": 4.473781388478582e-05,
      "loss": 0.0063,
      "step": 620
    },
    {
      "epoch": 0.23,
      "learning_rate": 4.472858197932053e-05,
      "loss": 0.14,
      "step": 621
    },
    {
      "epoch": 0.23,
      "learning_rate": 4.471935007385525e-05,
      "loss": 0.3122,
      "step": 622
    },
    {
      "epoch": 0.23,
      "learning_rate": 4.471011816838996e-05,
      "loss": 0.31,
      "step": 623
    },
    {
      "epoch": 0.23,
      "learning_rate": 4.470088626292467e-05,
      "loss": 0.0068,
      "step": 624
    },
    {
      "epoch": 0.23,
      "learning_rate": 4.469165435745938e-05,
      "loss": 0.2737,
      "step": 625
    },
    {
      "epoch": 0.23,
      "learning_rate": 4.468242245199409e-05,
      "loss": 0.5882,
      "step": 626
    },
    {
      "epoch": 0.23,
      "learning_rate": 4.46731905465288e-05,
      "loss": 0.189,
      "step": 627
    },
    {
      "epoch": 0.23,
      "learning_rate": 4.466395864106352e-05,
      "loss": 0.7588,
      "step": 628
    },
    {
      "epoch": 0.23,
      "learning_rate": 4.465472673559823e-05,
      "loss": 0.0665,
      "step": 629
    },
    {
      "epoch": 0.23,
      "learning_rate": 4.464549483013294e-05,
      "loss": 0.018,
      "step": 630
    },
    {
      "epoch": 0.23,
      "learning_rate": 4.463626292466765e-05,
      "loss": 0.7016,
      "step": 631
    },
    {
      "epoch": 0.23,
      "learning_rate": 4.462703101920236e-05,
      "loss": 0.0085,
      "step": 632
    },
    {
      "epoch": 0.23,
      "learning_rate": 4.461779911373708e-05,
      "loss": 0.3029,
      "step": 633
    },
    {
      "epoch": 0.23,
      "learning_rate": 4.460856720827179e-05,
      "loss": 0.0987,
      "step": 634
    },
    {
      "epoch": 0.23,
      "learning_rate": 4.45993353028065e-05,
      "loss": 0.4082,
      "step": 635
    },
    {
      "epoch": 0.23,
      "learning_rate": 4.459010339734122e-05,
      "loss": 0.1425,
      "step": 636
    },
    {
      "epoch": 0.23,
      "learning_rate": 4.458087149187593e-05,
      "loss": 0.0115,
      "step": 637
    },
    {
      "epoch": 0.23,
      "learning_rate": 4.457163958641063e-05,
      "loss": 0.3004,
      "step": 638
    },
    {
      "epoch": 0.23,
      "learning_rate": 4.456240768094535e-05,
      "loss": 0.57,
      "step": 639
    },
    {
      "epoch": 0.23,
      "learning_rate": 4.455317577548006e-05,
      "loss": 0.3387,
      "step": 640
    },
    {
      "epoch": 0.23,
      "learning_rate": 4.454394387001477e-05,
      "loss": 0.1401,
      "step": 641
    },
    {
      "epoch": 0.23,
      "learning_rate": 4.453471196454949e-05,
      "loss": 0.016,
      "step": 642
    },
    {
      "epoch": 0.24,
      "learning_rate": 4.45254800590842e-05,
      "loss": 0.0175,
      "step": 643
    },
    {
      "epoch": 0.24,
      "learning_rate": 4.451624815361891e-05,
      "loss": 0.0127,
      "step": 644
    },
    {
      "epoch": 0.24,
      "learning_rate": 4.450701624815362e-05,
      "loss": 0.0966,
      "step": 645
    },
    {
      "epoch": 0.24,
      "learning_rate": 4.449778434268833e-05,
      "loss": 0.1548,
      "step": 646
    },
    {
      "epoch": 0.24,
      "learning_rate": 4.448855243722305e-05,
      "loss": 0.2962,
      "step": 647
    },
    {
      "epoch": 0.24,
      "learning_rate": 4.447932053175776e-05,
      "loss": 0.5554,
      "step": 648
    },
    {
      "epoch": 0.24,
      "learning_rate": 4.447008862629247e-05,
      "loss": 0.2115,
      "step": 649
    },
    {
      "epoch": 0.24,
      "learning_rate": 4.446085672082718e-05,
      "loss": 0.1674,
      "step": 650
    },
    {
      "epoch": 0.24,
      "learning_rate": 4.445162481536189e-05,
      "loss": 0.4174,
      "step": 651
    },
    {
      "epoch": 0.24,
      "learning_rate": 4.444239290989661e-05,
      "loss": 0.2078,
      "step": 652
    },
    {
      "epoch": 0.24,
      "learning_rate": 4.443316100443132e-05,
      "loss": 0.03,
      "step": 653
    },
    {
      "epoch": 0.24,
      "learning_rate": 4.442392909896603e-05,
      "loss": 0.0257,
      "step": 654
    },
    {
      "epoch": 0.24,
      "learning_rate": 4.4414697193500745e-05,
      "loss": 0.0434,
      "step": 655
    },
    {
      "epoch": 0.24,
      "learning_rate": 4.4405465288035456e-05,
      "loss": 0.0326,
      "step": 656
    },
    {
      "epoch": 0.24,
      "learning_rate": 4.439623338257016e-05,
      "loss": 0.05,
      "step": 657
    },
    {
      "epoch": 0.24,
      "learning_rate": 4.438700147710488e-05,
      "loss": 0.1777,
      "step": 658
    },
    {
      "epoch": 0.24,
      "learning_rate": 4.437776957163959e-05,
      "loss": 0.0709,
      "step": 659
    },
    {
      "epoch": 0.24,
      "learning_rate": 4.43685376661743e-05,
      "loss": 0.017,
      "step": 660
    },
    {
      "epoch": 0.24,
      "learning_rate": 4.4359305760709015e-05,
      "loss": 0.3768,
      "step": 661
    },
    {
      "epoch": 0.24,
      "learning_rate": 4.4350073855243726e-05,
      "loss": 0.0145,
      "step": 662
    },
    {
      "epoch": 0.24,
      "learning_rate": 4.4340841949778436e-05,
      "loss": 0.1135,
      "step": 663
    },
    {
      "epoch": 0.24,
      "learning_rate": 4.433161004431315e-05,
      "loss": 0.0131,
      "step": 664
    },
    {
      "epoch": 0.24,
      "learning_rate": 4.432237813884786e-05,
      "loss": 0.0183,
      "step": 665
    },
    {
      "epoch": 0.24,
      "learning_rate": 4.4313146233382575e-05,
      "loss": 0.0081,
      "step": 666
    },
    {
      "epoch": 0.24,
      "learning_rate": 4.4303914327917285e-05,
      "loss": 1.1362,
      "step": 667
    },
    {
      "epoch": 0.24,
      "learning_rate": 4.4294682422451996e-05,
      "loss": 0.387,
      "step": 668
    },
    {
      "epoch": 0.24,
      "learning_rate": 4.4285450516986706e-05,
      "loss": 0.7142,
      "step": 669
    },
    {
      "epoch": 0.25,
      "learning_rate": 4.427621861152142e-05,
      "loss": 0.0154,
      "step": 670
    },
    {
      "epoch": 0.25,
      "learning_rate": 4.426698670605613e-05,
      "loss": 0.0137,
      "step": 671
    },
    {
      "epoch": 0.25,
      "learning_rate": 4.4257754800590845e-05,
      "loss": 0.1334,
      "step": 672
    },
    {
      "epoch": 0.25,
      "learning_rate": 4.4248522895125555e-05,
      "loss": 0.1497,
      "step": 673
    },
    {
      "epoch": 0.25,
      "learning_rate": 4.423929098966027e-05,
      "loss": 0.0158,
      "step": 674
    },
    {
      "epoch": 0.25,
      "learning_rate": 4.423005908419498e-05,
      "loss": 0.2056,
      "step": 675
    },
    {
      "epoch": 0.25,
      "learning_rate": 4.422082717872969e-05,
      "loss": 0.0673,
      "step": 676
    },
    {
      "epoch": 0.25,
      "learning_rate": 4.4211595273264404e-05,
      "loss": 0.1325,
      "step": 677
    },
    {
      "epoch": 0.25,
      "learning_rate": 4.4202363367799115e-05,
      "loss": 0.0126,
      "step": 678
    },
    {
      "epoch": 0.25,
      "learning_rate": 4.4193131462333825e-05,
      "loss": 0.0203,
      "step": 679
    },
    {
      "epoch": 0.25,
      "learning_rate": 4.418389955686854e-05,
      "loss": 0.075,
      "step": 680
    },
    {
      "epoch": 0.25,
      "learning_rate": 4.417466765140325e-05,
      "loss": 0.0051,
      "step": 681
    },
    {
      "epoch": 0.25,
      "learning_rate": 4.4165435745937964e-05,
      "loss": 0.0807,
      "step": 682
    },
    {
      "epoch": 0.25,
      "learning_rate": 4.4156203840472674e-05,
      "loss": 0.0098,
      "step": 683
    },
    {
      "epoch": 0.25,
      "learning_rate": 4.4146971935007385e-05,
      "loss": 0.0027,
      "step": 684
    },
    {
      "epoch": 0.25,
      "learning_rate": 4.41377400295421e-05,
      "loss": 0.3108,
      "step": 685
    },
    {
      "epoch": 0.25,
      "learning_rate": 4.412850812407681e-05,
      "loss": 0.0203,
      "step": 686
    },
    {
      "epoch": 0.25,
      "learning_rate": 4.411927621861152e-05,
      "loss": 0.0131,
      "step": 687
    },
    {
      "epoch": 0.25,
      "learning_rate": 4.4110044313146234e-05,
      "loss": 0.2893,
      "step": 688
    },
    {
      "epoch": 0.25,
      "learning_rate": 4.4100812407680944e-05,
      "loss": 0.656,
      "step": 689
    },
    {
      "epoch": 0.25,
      "learning_rate": 4.4091580502215655e-05,
      "loss": 0.0047,
      "step": 690
    },
    {
      "epoch": 0.25,
      "learning_rate": 4.408234859675037e-05,
      "loss": 0.0485,
      "step": 691
    },
    {
      "epoch": 0.25,
      "learning_rate": 4.407311669128508e-05,
      "loss": 0.3536,
      "step": 692
    },
    {
      "epoch": 0.25,
      "learning_rate": 4.40638847858198e-05,
      "loss": 0.1313,
      "step": 693
    },
    {
      "epoch": 0.25,
      "learning_rate": 4.405465288035451e-05,
      "loss": 0.0773,
      "step": 694
    },
    {
      "epoch": 0.25,
      "learning_rate": 4.4045420974889215e-05,
      "loss": 0.0068,
      "step": 695
    },
    {
      "epoch": 0.25,
      "learning_rate": 4.403618906942393e-05,
      "loss": 0.1755,
      "step": 696
    },
    {
      "epoch": 0.26,
      "learning_rate": 4.402695716395864e-05,
      "loss": 0.3711,
      "step": 697
    },
    {
      "epoch": 0.26,
      "learning_rate": 4.401772525849335e-05,
      "loss": 0.0043,
      "step": 698
    },
    {
      "epoch": 0.26,
      "learning_rate": 4.400849335302807e-05,
      "loss": 0.8702,
      "step": 699
    },
    {
      "epoch": 0.26,
      "learning_rate": 4.399926144756278e-05,
      "loss": 0.0126,
      "step": 700
    },
    {
      "epoch": 0.26,
      "learning_rate": 4.399002954209749e-05,
      "loss": 0.3116,
      "step": 701
    },
    {
      "epoch": 0.26,
      "learning_rate": 4.39807976366322e-05,
      "loss": 0.0062,
      "step": 702
    },
    {
      "epoch": 0.26,
      "learning_rate": 4.397156573116691e-05,
      "loss": 0.0036,
      "step": 703
    },
    {
      "epoch": 0.26,
      "learning_rate": 4.396233382570163e-05,
      "loss": 0.3133,
      "step": 704
    },
    {
      "epoch": 0.26,
      "learning_rate": 4.395310192023634e-05,
      "loss": 0.0079,
      "step": 705
    },
    {
      "epoch": 0.26,
      "learning_rate": 4.394387001477105e-05,
      "loss": 0.1102,
      "step": 706
    },
    {
      "epoch": 0.26,
      "learning_rate": 4.393463810930576e-05,
      "loss": 0.5496,
      "step": 707
    },
    {
      "epoch": 0.26,
      "learning_rate": 4.392540620384047e-05,
      "loss": 0.3995,
      "step": 708
    },
    {
      "epoch": 0.26,
      "learning_rate": 4.391617429837518e-05,
      "loss": 0.2206,
      "step": 709
    },
    {
      "epoch": 0.26,
      "learning_rate": 4.39069423929099e-05,
      "loss": 0.0462,
      "step": 710
    },
    {
      "epoch": 0.26,
      "learning_rate": 4.389771048744461e-05,
      "loss": 0.0224,
      "step": 711
    },
    {
      "epoch": 0.26,
      "learning_rate": 4.388847858197933e-05,
      "loss": 0.0412,
      "step": 712
    },
    {
      "epoch": 0.26,
      "learning_rate": 4.387924667651404e-05,
      "loss": 0.0124,
      "step": 713
    },
    {
      "epoch": 0.26,
      "learning_rate": 4.387001477104874e-05,
      "loss": 0.1225,
      "step": 714
    },
    {
      "epoch": 0.26,
      "learning_rate": 4.386078286558346e-05,
      "loss": 0.2149,
      "step": 715
    },
    {
      "epoch": 0.26,
      "learning_rate": 4.385155096011817e-05,
      "loss": 0.0772,
      "step": 716
    },
    {
      "epoch": 0.26,
      "learning_rate": 4.384231905465288e-05,
      "loss": 0.2448,
      "step": 717
    },
    {
      "epoch": 0.26,
      "learning_rate": 4.38330871491876e-05,
      "loss": 0.0569,
      "step": 718
    },
    {
      "epoch": 0.26,
      "learning_rate": 4.382385524372231e-05,
      "loss": 0.6738,
      "step": 719
    },
    {
      "epoch": 0.26,
      "learning_rate": 4.381462333825702e-05,
      "loss": 0.0132,
      "step": 720
    },
    {
      "epoch": 0.26,
      "learning_rate": 4.380539143279173e-05,
      "loss": 0.013,
      "step": 721
    },
    {
      "epoch": 0.26,
      "learning_rate": 4.379615952732644e-05,
      "loss": 0.2893,
      "step": 722
    },
    {
      "epoch": 0.26,
      "learning_rate": 4.378692762186116e-05,
      "loss": 0.0166,
      "step": 723
    },
    {
      "epoch": 0.26,
      "learning_rate": 4.377769571639587e-05,
      "loss": 0.0679,
      "step": 724
    },
    {
      "epoch": 0.27,
      "learning_rate": 4.376846381093058e-05,
      "loss": 0.4278,
      "step": 725
    },
    {
      "epoch": 0.27,
      "learning_rate": 4.375923190546529e-05,
      "loss": 0.0852,
      "step": 726
    },
    {
      "epoch": 0.27,
      "learning_rate": 4.375e-05,
      "loss": 0.0303,
      "step": 727
    },
    {
      "epoch": 0.27,
      "learning_rate": 4.374076809453471e-05,
      "loss": 0.0359,
      "step": 728
    },
    {
      "epoch": 0.27,
      "learning_rate": 4.373153618906943e-05,
      "loss": 0.4899,
      "step": 729
    },
    {
      "epoch": 0.27,
      "learning_rate": 4.372230428360414e-05,
      "loss": 0.0395,
      "step": 730
    },
    {
      "epoch": 0.27,
      "learning_rate": 4.371307237813885e-05,
      "loss": 0.0671,
      "step": 731
    },
    {
      "epoch": 0.27,
      "learning_rate": 4.3703840472673566e-05,
      "loss": 0.0455,
      "step": 732
    },
    {
      "epoch": 0.27,
      "learning_rate": 4.369460856720827e-05,
      "loss": 0.023,
      "step": 733
    },
    {
      "epoch": 0.27,
      "learning_rate": 4.368537666174299e-05,
      "loss": 0.4225,
      "step": 734
    },
    {
      "epoch": 0.27,
      "learning_rate": 4.36761447562777e-05,
      "loss": 0.0318,
      "step": 735
    },
    {
      "epoch": 0.27,
      "learning_rate": 4.366691285081241e-05,
      "loss": 0.434,
      "step": 736
    },
    {
      "epoch": 0.27,
      "learning_rate": 4.3657680945347125e-05,
      "loss": 0.0237,
      "step": 737
    },
    {
      "epoch": 0.27,
      "learning_rate": 4.3648449039881836e-05,
      "loss": 0.0817,
      "step": 738
    },
    {
      "epoch": 0.27,
      "learning_rate": 4.3639217134416546e-05,
      "loss": 0.2249,
      "step": 739
    },
    {
      "epoch": 0.27,
      "learning_rate": 4.362998522895126e-05,
      "loss": 0.0629,
      "step": 740
    },
    {
      "epoch": 0.27,
      "learning_rate": 4.362075332348597e-05,
      "loss": 0.0093,
      "step": 741
    },
    {
      "epoch": 0.27,
      "learning_rate": 4.361152141802068e-05,
      "loss": 0.0071,
      "step": 742
    },
    {
      "epoch": 0.27,
      "learning_rate": 4.3602289512555395e-05,
      "loss": 0.0287,
      "step": 743
    },
    {
      "epoch": 0.27,
      "learning_rate": 4.3593057607090106e-05,
      "loss": 0.2444,
      "step": 744
    },
    {
      "epoch": 0.27,
      "learning_rate": 4.3583825701624817e-05,
      "loss": 0.127,
      "step": 745
    },
    {
      "epoch": 0.27,
      "learning_rate": 4.357459379615953e-05,
      "loss": 0.1628,
      "step": 746
    },
    {
      "epoch": 0.27,
      "learning_rate": 4.356536189069424e-05,
      "loss": 0.0148,
      "step": 747
    },
    {
      "epoch": 0.27,
      "learning_rate": 4.3556129985228955e-05,
      "loss": 0.0382,
      "step": 748
    },
    {
      "epoch": 0.27,
      "learning_rate": 4.3546898079763666e-05,
      "loss": 0.0165,
      "step": 749
    },
    {
      "epoch": 0.27,
      "learning_rate": 4.3537666174298376e-05,
      "loss": 0.1825,
      "step": 750
    },
    {
      "epoch": 0.27,
      "learning_rate": 4.3528434268833093e-05,
      "loss": 0.5154,
      "step": 751
    },
    {
      "epoch": 0.28,
      "learning_rate": 4.35192023633678e-05,
      "loss": 0.4209,
      "step": 752
    },
    {
      "epoch": 0.28,
      "learning_rate": 4.3509970457902514e-05,
      "loss": 0.3283,
      "step": 753
    },
    {
      "epoch": 0.28,
      "learning_rate": 4.3500738552437225e-05,
      "loss": 0.3475,
      "step": 754
    },
    {
      "epoch": 0.28,
      "learning_rate": 4.3491506646971936e-05,
      "loss": 0.0391,
      "step": 755
    },
    {
      "epoch": 0.28,
      "learning_rate": 4.348227474150665e-05,
      "loss": 0.4097,
      "step": 756
    },
    {
      "epoch": 0.28,
      "learning_rate": 4.3473042836041363e-05,
      "loss": 0.0034,
      "step": 757
    },
    {
      "epoch": 0.28,
      "learning_rate": 4.3463810930576074e-05,
      "loss": 0.452,
      "step": 758
    },
    {
      "epoch": 0.28,
      "learning_rate": 4.3454579025110785e-05,
      "loss": 0.0634,
      "step": 759
    },
    {
      "epoch": 0.28,
      "learning_rate": 4.3445347119645495e-05,
      "loss": 0.1668,
      "step": 760
    },
    {
      "epoch": 0.28,
      "learning_rate": 4.3436115214180206e-05,
      "loss": 0.1573,
      "step": 761
    },
    {
      "epoch": 0.28,
      "learning_rate": 4.342688330871492e-05,
      "loss": 0.1076,
      "step": 762
    },
    {
      "epoch": 0.28,
      "learning_rate": 4.3417651403249634e-05,
      "loss": 0.0948,
      "step": 763
    },
    {
      "epoch": 0.28,
      "learning_rate": 4.3408419497784344e-05,
      "loss": 0.2209,
      "step": 764
    },
    {
      "epoch": 0.28,
      "learning_rate": 4.3399187592319055e-05,
      "loss": 0.2842,
      "step": 765
    },
    {
      "epoch": 0.28,
      "learning_rate": 4.3389955686853765e-05,
      "loss": 0.0927,
      "step": 766
    },
    {
      "epoch": 0.28,
      "learning_rate": 4.338072378138848e-05,
      "loss": 0.4353,
      "step": 767
    },
    {
      "epoch": 0.28,
      "learning_rate": 4.337149187592319e-05,
      "loss": 0.033,
      "step": 768
    },
    {
      "epoch": 0.28,
      "learning_rate": 4.3362259970457904e-05,
      "loss": 0.2534,
      "step": 769
    },
    {
      "epoch": 0.28,
      "learning_rate": 4.335302806499262e-05,
      "loss": 0.0201,
      "step": 770
    },
    {
      "epoch": 0.28,
      "learning_rate": 4.3343796159527325e-05,
      "loss": 0.2762,
      "step": 771
    },
    {
      "epoch": 0.28,
      "learning_rate": 4.3334564254062035e-05,
      "loss": 0.0975,
      "step": 772
    },
    {
      "epoch": 0.28,
      "learning_rate": 4.332533234859675e-05,
      "loss": 0.3521,
      "step": 773
    },
    {
      "epoch": 0.28,
      "learning_rate": 4.331610044313146e-05,
      "loss": 0.0132,
      "step": 774
    },
    {
      "epoch": 0.28,
      "learning_rate": 4.330686853766618e-05,
      "loss": 0.074,
      "step": 775
    },
    {
      "epoch": 0.28,
      "learning_rate": 4.329763663220089e-05,
      "loss": 0.0231,
      "step": 776
    },
    {
      "epoch": 0.28,
      "learning_rate": 4.32884047267356e-05,
      "loss": 0.0551,
      "step": 777
    },
    {
      "epoch": 0.28,
      "learning_rate": 4.327917282127031e-05,
      "loss": 0.0379,
      "step": 778
    },
    {
      "epoch": 0.29,
      "learning_rate": 4.326994091580502e-05,
      "loss": 0.0777,
      "step": 779
    },
    {
      "epoch": 0.29,
      "learning_rate": 4.326070901033973e-05,
      "loss": 0.0156,
      "step": 780
    },
    {
      "epoch": 0.29,
      "learning_rate": 4.325147710487445e-05,
      "loss": 0.2059,
      "step": 781
    },
    {
      "epoch": 0.29,
      "learning_rate": 4.324224519940916e-05,
      "loss": 0.1033,
      "step": 782
    },
    {
      "epoch": 0.29,
      "learning_rate": 4.323301329394387e-05,
      "loss": 0.0188,
      "step": 783
    },
    {
      "epoch": 0.29,
      "learning_rate": 4.322378138847858e-05,
      "loss": 0.0397,
      "step": 784
    },
    {
      "epoch": 0.29,
      "learning_rate": 4.321454948301329e-05,
      "loss": 0.1153,
      "step": 785
    },
    {
      "epoch": 0.29,
      "learning_rate": 4.320531757754801e-05,
      "loss": 0.0731,
      "step": 786
    },
    {
      "epoch": 0.29,
      "learning_rate": 4.319608567208272e-05,
      "loss": 0.037,
      "step": 787
    },
    {
      "epoch": 0.29,
      "learning_rate": 4.318685376661743e-05,
      "loss": 0.2293,
      "step": 788
    },
    {
      "epoch": 0.29,
      "learning_rate": 4.317762186115215e-05,
      "loss": 0.0056,
      "step": 789
    },
    {
      "epoch": 0.29,
      "learning_rate": 4.316838995568685e-05,
      "loss": 0.0117,
      "step": 790
    },
    {
      "epoch": 0.29,
      "learning_rate": 4.315915805022156e-05,
      "loss": 0.1594,
      "step": 791
    },
    {
      "epoch": 0.29,
      "learning_rate": 4.314992614475628e-05,
      "loss": 0.0047,
      "step": 792
    },
    {
      "epoch": 0.29,
      "learning_rate": 4.314069423929099e-05,
      "loss": 0.0046,
      "step": 793
    },
    {
      "epoch": 0.29,
      "learning_rate": 4.313146233382571e-05,
      "loss": 0.0045,
      "step": 794
    },
    {
      "epoch": 0.29,
      "learning_rate": 4.312223042836042e-05,
      "loss": 0.0039,
      "step": 795
    },
    {
      "epoch": 0.29,
      "learning_rate": 4.311299852289513e-05,
      "loss": 0.2916,
      "step": 796
    },
    {
      "epoch": 0.29,
      "learning_rate": 4.310376661742984e-05,
      "loss": 0.2623,
      "step": 797
    },
    {
      "epoch": 0.29,
      "learning_rate": 4.309453471196455e-05,
      "loss": 0.1563,
      "step": 798
    },
    {
      "epoch": 0.29,
      "learning_rate": 4.308530280649926e-05,
      "loss": 0.0337,
      "step": 799
    },
    {
      "epoch": 0.29,
      "learning_rate": 4.307607090103398e-05,
      "loss": 0.0047,
      "step": 800
    },
    {
      "epoch": 0.29,
      "learning_rate": 4.306683899556869e-05,
      "loss": 0.5374,
      "step": 801
    },
    {
      "epoch": 0.29,
      "learning_rate": 4.30576070901034e-05,
      "loss": 0.0047,
      "step": 802
    },
    {
      "epoch": 0.29,
      "learning_rate": 4.304837518463811e-05,
      "loss": 0.3362,
      "step": 803
    },
    {
      "epoch": 0.29,
      "learning_rate": 4.303914327917282e-05,
      "loss": 0.002,
      "step": 804
    },
    {
      "epoch": 0.29,
      "learning_rate": 4.302991137370754e-05,
      "loss": 0.2808,
      "step": 805
    },
    {
      "epoch": 0.29,
      "learning_rate": 4.302067946824225e-05,
      "loss": 0.0609,
      "step": 806
    },
    {
      "epoch": 0.3,
      "learning_rate": 4.301144756277696e-05,
      "loss": 0.2006,
      "step": 807
    },
    {
      "epoch": 0.3,
      "learning_rate": 4.3002215657311676e-05,
      "loss": 0.0918,
      "step": 808
    },
    {
      "epoch": 0.3,
      "learning_rate": 4.299298375184638e-05,
      "loss": 0.1608,
      "step": 809
    },
    {
      "epoch": 0.3,
      "learning_rate": 4.298375184638109e-05,
      "loss": 0.0369,
      "step": 810
    },
    {
      "epoch": 0.3,
      "learning_rate": 4.297451994091581e-05,
      "loss": 0.1412,
      "step": 811
    },
    {
      "epoch": 0.3,
      "learning_rate": 4.296528803545052e-05,
      "loss": 0.3153,
      "step": 812
    },
    {
      "epoch": 0.3,
      "learning_rate": 4.2956056129985236e-05,
      "loss": 0.5814,
      "step": 813
    },
    {
      "epoch": 0.3,
      "learning_rate": 4.2946824224519946e-05,
      "loss": 0.0073,
      "step": 814
    },
    {
      "epoch": 0.3,
      "learning_rate": 4.293759231905466e-05,
      "loss": 0.0069,
      "step": 815
    },
    {
      "epoch": 0.3,
      "learning_rate": 4.292836041358937e-05,
      "loss": 0.0786,
      "step": 816
    },
    {
      "epoch": 0.3,
      "learning_rate": 4.291912850812408e-05,
      "loss": 0.2602,
      "step": 817
    },
    {
      "epoch": 0.3,
      "learning_rate": 4.290989660265879e-05,
      "loss": 0.0074,
      "step": 818
    },
    {
      "epoch": 0.3,
      "learning_rate": 4.2900664697193506e-05,
      "loss": 0.0185,
      "step": 819
    },
    {
      "epoch": 0.3,
      "learning_rate": 4.2891432791728216e-05,
      "loss": 0.0209,
      "step": 820
    },
    {
      "epoch": 0.3,
      "learning_rate": 4.288220088626293e-05,
      "loss": 0.0126,
      "step": 821
    },
    {
      "epoch": 0.3,
      "learning_rate": 4.287296898079764e-05,
      "loss": 0.0185,
      "step": 822
    },
    {
      "epoch": 0.3,
      "learning_rate": 4.286373707533235e-05,
      "loss": 0.01,
      "step": 823
    },
    {
      "epoch": 0.3,
      "learning_rate": 4.2854505169867065e-05,
      "loss": 0.459,
      "step": 824
    },
    {
      "epoch": 0.3,
      "learning_rate": 4.2845273264401776e-05,
      "loss": 0.0039,
      "step": 825
    },
    {
      "epoch": 0.3,
      "learning_rate": 4.2836041358936486e-05,
      "loss": 0.02,
      "step": 826
    },
    {
      "epoch": 0.3,
      "learning_rate": 4.2826809453471204e-05,
      "loss": 0.0048,
      "step": 827
    },
    {
      "epoch": 0.3,
      "learning_rate": 4.281757754800591e-05,
      "loss": 0.1054,
      "step": 828
    },
    {
      "epoch": 0.3,
      "learning_rate": 4.280834564254062e-05,
      "loss": 0.4372,
      "step": 829
    },
    {
      "epoch": 0.3,
      "learning_rate": 4.2799113737075335e-05,
      "loss": 0.0048,
      "step": 830
    },
    {
      "epoch": 0.3,
      "learning_rate": 4.2789881831610046e-05,
      "loss": 0.0048,
      "step": 831
    },
    {
      "epoch": 0.3,
      "learning_rate": 4.2780649926144756e-05,
      "loss": 0.2683,
      "step": 832
    },
    {
      "epoch": 0.3,
      "learning_rate": 4.2771418020679474e-05,
      "loss": 0.003,
      "step": 833
    },
    {
      "epoch": 0.31,
      "learning_rate": 4.2762186115214184e-05,
      "loss": 0.0024,
      "step": 834
    },
    {
      "epoch": 0.31,
      "learning_rate": 4.2752954209748895e-05,
      "loss": 0.3914,
      "step": 835
    },
    {
      "epoch": 0.31,
      "learning_rate": 4.2743722304283605e-05,
      "loss": 0.1135,
      "step": 836
    },
    {
      "epoch": 0.31,
      "learning_rate": 4.2734490398818316e-05,
      "loss": 0.0024,
      "step": 837
    },
    {
      "epoch": 0.31,
      "learning_rate": 4.272525849335303e-05,
      "loss": 0.2688,
      "step": 838
    },
    {
      "epoch": 0.31,
      "learning_rate": 4.2716026587887744e-05,
      "loss": 0.0025,
      "step": 839
    },
    {
      "epoch": 0.31,
      "learning_rate": 4.2706794682422454e-05,
      "loss": 0.0207,
      "step": 840
    },
    {
      "epoch": 0.31,
      "learning_rate": 4.2697562776957165e-05,
      "loss": 0.4197,
      "step": 841
    },
    {
      "epoch": 0.31,
      "learning_rate": 4.2688330871491875e-05,
      "loss": 0.0539,
      "step": 842
    },
    {
      "epoch": 0.31,
      "learning_rate": 4.2679098966026586e-05,
      "loss": 0.2595,
      "step": 843
    },
    {
      "epoch": 0.31,
      "learning_rate": 4.26698670605613e-05,
      "loss": 0.2817,
      "step": 844
    },
    {
      "epoch": 0.31,
      "learning_rate": 4.2660635155096014e-05,
      "loss": 0.4004,
      "step": 845
    },
    {
      "epoch": 0.31,
      "learning_rate": 4.265140324963073e-05,
      "loss": 0.0323,
      "step": 846
    },
    {
      "epoch": 0.31,
      "learning_rate": 4.2642171344165435e-05,
      "loss": 0.6108,
      "step": 847
    },
    {
      "epoch": 0.31,
      "learning_rate": 4.2632939438700145e-05,
      "loss": 0.3753,
      "step": 848
    },
    {
      "epoch": 0.31,
      "learning_rate": 4.262370753323486e-05,
      "loss": 0.2176,
      "step": 849
    },
    {
      "epoch": 0.31,
      "learning_rate": 4.261447562776957e-05,
      "loss": 0.4671,
      "step": 850
    },
    {
      "epoch": 0.31,
      "learning_rate": 4.2605243722304284e-05,
      "loss": 0.0276,
      "step": 851
    },
    {
      "epoch": 0.31,
      "learning_rate": 4.2596011816839e-05,
      "loss": 0.2363,
      "step": 852
    },
    {
      "epoch": 0.31,
      "learning_rate": 4.258677991137371e-05,
      "loss": 0.2303,
      "step": 853
    },
    {
      "epoch": 0.31,
      "learning_rate": 4.257754800590842e-05,
      "loss": 0.0055,
      "step": 854
    },
    {
      "epoch": 0.31,
      "learning_rate": 4.256831610044313e-05,
      "loss": 0.0897,
      "step": 855
    },
    {
      "epoch": 0.31,
      "learning_rate": 4.255908419497784e-05,
      "loss": 0.3951,
      "step": 856
    },
    {
      "epoch": 0.31,
      "learning_rate": 4.254985228951256e-05,
      "loss": 0.0228,
      "step": 857
    },
    {
      "epoch": 0.31,
      "learning_rate": 4.254062038404727e-05,
      "loss": 0.2288,
      "step": 858
    },
    {
      "epoch": 0.31,
      "learning_rate": 4.253138847858198e-05,
      "loss": 0.0169,
      "step": 859
    },
    {
      "epoch": 0.31,
      "learning_rate": 4.252215657311669e-05,
      "loss": 0.2136,
      "step": 860
    },
    {
      "epoch": 0.32,
      "learning_rate": 4.25129246676514e-05,
      "loss": 0.0084,
      "step": 861
    },
    {
      "epoch": 0.32,
      "learning_rate": 4.2503692762186113e-05,
      "loss": 0.0285,
      "step": 862
    },
    {
      "epoch": 0.32,
      "learning_rate": 4.249446085672083e-05,
      "loss": 0.0152,
      "step": 863
    },
    {
      "epoch": 0.32,
      "learning_rate": 4.248522895125554e-05,
      "loss": 0.0052,
      "step": 864
    },
    {
      "epoch": 0.32,
      "learning_rate": 4.247599704579026e-05,
      "loss": 0.3927,
      "step": 865
    },
    {
      "epoch": 0.32,
      "learning_rate": 4.246676514032496e-05,
      "loss": 0.9152,
      "step": 866
    },
    {
      "epoch": 0.32,
      "learning_rate": 4.245753323485967e-05,
      "loss": 0.0066,
      "step": 867
    },
    {
      "epoch": 0.32,
      "learning_rate": 4.244830132939439e-05,
      "loss": 0.0346,
      "step": 868
    },
    {
      "epoch": 0.32,
      "learning_rate": 4.24390694239291e-05,
      "loss": 0.3752,
      "step": 869
    },
    {
      "epoch": 0.32,
      "learning_rate": 4.242983751846381e-05,
      "loss": 0.0086,
      "step": 870
    },
    {
      "epoch": 0.32,
      "learning_rate": 4.242060561299853e-05,
      "loss": 0.0201,
      "step": 871
    },
    {
      "epoch": 0.32,
      "learning_rate": 4.241137370753324e-05,
      "loss": 0.6786,
      "step": 872
    },
    {
      "epoch": 0.32,
      "learning_rate": 4.240214180206794e-05,
      "loss": 0.5408,
      "step": 873
    },
    {
      "epoch": 0.32,
      "learning_rate": 4.239290989660266e-05,
      "loss": 0.4389,
      "step": 874
    },
    {
      "epoch": 0.32,
      "learning_rate": 4.238367799113737e-05,
      "loss": 0.0408,
      "step": 875
    },
    {
      "epoch": 0.32,
      "learning_rate": 4.237444608567209e-05,
      "loss": 0.0674,
      "step": 876
    },
    {
      "epoch": 0.32,
      "learning_rate": 4.23652141802068e-05,
      "loss": 0.0073,
      "step": 877
    },
    {
      "epoch": 0.32,
      "learning_rate": 4.235598227474151e-05,
      "loss": 0.3847,
      "step": 878
    },
    {
      "epoch": 0.32,
      "learning_rate": 4.234675036927622e-05,
      "loss": 0.0057,
      "step": 879
    },
    {
      "epoch": 0.32,
      "learning_rate": 4.233751846381093e-05,
      "loss": 0.0664,
      "step": 880
    },
    {
      "epoch": 0.32,
      "learning_rate": 4.232828655834564e-05,
      "loss": 0.0402,
      "step": 881
    },
    {
      "epoch": 0.32,
      "learning_rate": 4.231905465288036e-05,
      "loss": 0.302,
      "step": 882
    },
    {
      "epoch": 0.32,
      "learning_rate": 4.230982274741507e-05,
      "loss": 0.0973,
      "step": 883
    },
    {
      "epoch": 0.32,
      "learning_rate": 4.2300590841949786e-05,
      "loss": 0.2348,
      "step": 884
    },
    {
      "epoch": 0.32,
      "learning_rate": 4.229135893648449e-05,
      "loss": 0.0102,
      "step": 885
    },
    {
      "epoch": 0.32,
      "learning_rate": 4.22821270310192e-05,
      "loss": 0.152,
      "step": 886
    },
    {
      "epoch": 0.32,
      "learning_rate": 4.227289512555392e-05,
      "loss": 0.3155,
      "step": 887
    },
    {
      "epoch": 0.32,
      "learning_rate": 4.226366322008863e-05,
      "loss": 0.007,
      "step": 888
    },
    {
      "epoch": 0.33,
      "learning_rate": 4.225443131462334e-05,
      "loss": 0.0903,
      "step": 889
    },
    {
      "epoch": 0.33,
      "learning_rate": 4.2245199409158056e-05,
      "loss": 0.1586,
      "step": 890
    },
    {
      "epoch": 0.33,
      "learning_rate": 4.223596750369277e-05,
      "loss": 0.0139,
      "step": 891
    },
    {
      "epoch": 0.33,
      "learning_rate": 4.222673559822747e-05,
      "loss": 0.0114,
      "step": 892
    },
    {
      "epoch": 0.33,
      "learning_rate": 4.221750369276219e-05,
      "loss": 0.25,
      "step": 893
    },
    {
      "epoch": 0.33,
      "learning_rate": 4.22082717872969e-05,
      "loss": 0.0785,
      "step": 894
    },
    {
      "epoch": 0.33,
      "learning_rate": 4.2199039881831616e-05,
      "loss": 0.0057,
      "step": 895
    },
    {
      "epoch": 0.33,
      "learning_rate": 4.2189807976366326e-05,
      "loss": 0.013,
      "step": 896
    },
    {
      "epoch": 0.33,
      "learning_rate": 4.218057607090104e-05,
      "loss": 0.2226,
      "step": 897
    },
    {
      "epoch": 0.33,
      "learning_rate": 4.217134416543575e-05,
      "loss": 0.1045,
      "step": 898
    },
    {
      "epoch": 0.33,
      "learning_rate": 4.216211225997046e-05,
      "loss": 0.2249,
      "step": 899
    },
    {
      "epoch": 0.33,
      "learning_rate": 4.215288035450517e-05,
      "loss": 0.0075,
      "step": 900
    },
    {
      "epoch": 0.33,
      "learning_rate": 4.2143648449039886e-05,
      "loss": 0.0625,
      "step": 901
    },
    {
      "epoch": 0.33,
      "learning_rate": 4.2134416543574596e-05,
      "loss": 0.02,
      "step": 902
    },
    {
      "epoch": 0.33,
      "learning_rate": 4.212518463810931e-05,
      "loss": 0.004,
      "step": 903
    },
    {
      "epoch": 0.33,
      "learning_rate": 4.211595273264402e-05,
      "loss": 0.1215,
      "step": 904
    },
    {
      "epoch": 0.33,
      "learning_rate": 4.210672082717873e-05,
      "loss": 0.0039,
      "step": 905
    },
    {
      "epoch": 0.33,
      "learning_rate": 4.2097488921713445e-05,
      "loss": 0.155,
      "step": 906
    },
    {
      "epoch": 0.33,
      "learning_rate": 4.2088257016248156e-05,
      "loss": 0.002,
      "step": 907
    },
    {
      "epoch": 0.33,
      "learning_rate": 4.2079025110782866e-05,
      "loss": 0.0753,
      "step": 908
    },
    {
      "epoch": 0.33,
      "learning_rate": 4.2069793205317584e-05,
      "loss": 0.2478,
      "step": 909
    },
    {
      "epoch": 0.33,
      "learning_rate": 4.2060561299852294e-05,
      "loss": 0.3115,
      "step": 910
    },
    {
      "epoch": 0.33,
      "learning_rate": 4.2051329394387e-05,
      "loss": 0.0874,
      "step": 911
    },
    {
      "epoch": 0.33,
      "learning_rate": 4.2042097488921715e-05,
      "loss": 0.0065,
      "step": 912
    },
    {
      "epoch": 0.33,
      "learning_rate": 4.2032865583456426e-05,
      "loss": 0.0149,
      "step": 913
    },
    {
      "epoch": 0.33,
      "learning_rate": 4.202363367799114e-05,
      "loss": 0.6069,
      "step": 914
    },
    {
      "epoch": 0.33,
      "learning_rate": 4.2014401772525854e-05,
      "loss": 0.0633,
      "step": 915
    },
    {
      "epoch": 0.34,
      "learning_rate": 4.2005169867060564e-05,
      "loss": 0.0053,
      "step": 916
    },
    {
      "epoch": 0.34,
      "learning_rate": 4.1995937961595275e-05,
      "loss": 0.399,
      "step": 917
    },
    {
      "epoch": 0.34,
      "learning_rate": 4.1986706056129985e-05,
      "loss": 0.2964,
      "step": 918
    },
    {
      "epoch": 0.34,
      "learning_rate": 4.1977474150664696e-05,
      "loss": 0.0618,
      "step": 919
    },
    {
      "epoch": 0.34,
      "learning_rate": 4.196824224519941e-05,
      "loss": 0.1145,
      "step": 920
    },
    {
      "epoch": 0.34,
      "learning_rate": 4.1959010339734124e-05,
      "loss": 0.0207,
      "step": 921
    },
    {
      "epoch": 0.34,
      "learning_rate": 4.1949778434268834e-05,
      "loss": 0.276,
      "step": 922
    },
    {
      "epoch": 0.34,
      "learning_rate": 4.1940546528803545e-05,
      "loss": 0.009,
      "step": 923
    },
    {
      "epoch": 0.34,
      "learning_rate": 4.1931314623338256e-05,
      "loss": 0.4704,
      "step": 924
    },
    {
      "epoch": 0.34,
      "learning_rate": 4.192208271787297e-05,
      "loss": 0.1054,
      "step": 925
    },
    {
      "epoch": 0.34,
      "learning_rate": 4.1912850812407683e-05,
      "loss": 0.0268,
      "step": 926
    },
    {
      "epoch": 0.34,
      "learning_rate": 4.1903618906942394e-05,
      "loss": 0.0201,
      "step": 927
    },
    {
      "epoch": 0.34,
      "learning_rate": 4.189438700147711e-05,
      "loss": 0.1226,
      "step": 928
    },
    {
      "epoch": 0.34,
      "learning_rate": 4.188515509601182e-05,
      "loss": 0.3249,
      "step": 929
    },
    {
      "epoch": 0.34,
      "learning_rate": 4.1875923190546526e-05,
      "loss": 0.0523,
      "step": 930
    },
    {
      "epoch": 0.34,
      "learning_rate": 4.186669128508124e-05,
      "loss": 0.0524,
      "step": 931
    },
    {
      "epoch": 0.34,
      "learning_rate": 4.1857459379615953e-05,
      "loss": 0.1611,
      "step": 932
    },
    {
      "epoch": 0.34,
      "learning_rate": 4.1848227474150664e-05,
      "loss": 0.1537,
      "step": 933
    },
    {
      "epoch": 0.34,
      "learning_rate": 4.183899556868538e-05,
      "loss": 0.2531,
      "step": 934
    },
    {
      "epoch": 0.34,
      "learning_rate": 4.182976366322009e-05,
      "loss": 0.0835,
      "step": 935
    },
    {
      "epoch": 0.34,
      "learning_rate": 4.18205317577548e-05,
      "loss": 0.0175,
      "step": 936
    },
    {
      "epoch": 0.34,
      "learning_rate": 4.181129985228951e-05,
      "loss": 0.0579,
      "step": 937
    },
    {
      "epoch": 0.34,
      "learning_rate": 4.1802067946824224e-05,
      "loss": 0.0092,
      "step": 938
    },
    {
      "epoch": 0.34,
      "learning_rate": 4.179283604135894e-05,
      "loss": 0.2586,
      "step": 939
    },
    {
      "epoch": 0.34,
      "learning_rate": 4.178360413589365e-05,
      "loss": 0.0979,
      "step": 940
    },
    {
      "epoch": 0.34,
      "learning_rate": 4.177437223042836e-05,
      "loss": 0.0265,
      "step": 941
    },
    {
      "epoch": 0.34,
      "learning_rate": 4.176514032496307e-05,
      "loss": 0.3937,
      "step": 942
    },
    {
      "epoch": 0.35,
      "learning_rate": 4.175590841949778e-05,
      "loss": 0.2698,
      "step": 943
    },
    {
      "epoch": 0.35,
      "learning_rate": 4.1746676514032494e-05,
      "loss": 0.0411,
      "step": 944
    },
    {
      "epoch": 0.35,
      "learning_rate": 4.173744460856721e-05,
      "loss": 0.1673,
      "step": 945
    },
    {
      "epoch": 0.35,
      "learning_rate": 4.172821270310192e-05,
      "loss": 0.0168,
      "step": 946
    },
    {
      "epoch": 0.35,
      "learning_rate": 4.171898079763664e-05,
      "loss": 0.026,
      "step": 947
    },
    {
      "epoch": 0.35,
      "learning_rate": 4.170974889217135e-05,
      "loss": 0.042,
      "step": 948
    },
    {
      "epoch": 0.35,
      "learning_rate": 4.170051698670605e-05,
      "loss": 0.0393,
      "step": 949
    },
    {
      "epoch": 0.35,
      "learning_rate": 4.169128508124077e-05,
      "loss": 0.2036,
      "step": 950
    },
    {
      "epoch": 0.35,
      "learning_rate": 4.168205317577548e-05,
      "loss": 0.0194,
      "step": 951
    },
    {
      "epoch": 0.35,
      "learning_rate": 4.167282127031019e-05,
      "loss": 0.2968,
      "step": 952
    },
    {
      "epoch": 0.35,
      "learning_rate": 4.166358936484491e-05,
      "loss": 0.0376,
      "step": 953
    },
    {
      "epoch": 0.35,
      "learning_rate": 4.165435745937962e-05,
      "loss": 0.0038,
      "step": 954
    },
    {
      "epoch": 0.35,
      "learning_rate": 4.164512555391433e-05,
      "loss": 0.0827,
      "step": 955
    },
    {
      "epoch": 0.35,
      "learning_rate": 4.163589364844904e-05,
      "loss": 0.0127,
      "step": 956
    },
    {
      "epoch": 0.35,
      "learning_rate": 4.162666174298375e-05,
      "loss": 0.2226,
      "step": 957
    },
    {
      "epoch": 0.35,
      "learning_rate": 4.161742983751847e-05,
      "loss": 0.5884,
      "step": 958
    },
    {
      "epoch": 0.35,
      "learning_rate": 4.160819793205318e-05,
      "loss": 0.2176,
      "step": 959
    },
    {
      "epoch": 0.35,
      "learning_rate": 4.159896602658789e-05,
      "loss": 0.1487,
      "step": 960
    },
    {
      "epoch": 0.35,
      "learning_rate": 4.15897341211226e-05,
      "loss": 0.3638,
      "step": 961
    },
    {
      "epoch": 0.35,
      "learning_rate": 4.158050221565731e-05,
      "loss": 0.1128,
      "step": 962
    },
    {
      "epoch": 0.35,
      "learning_rate": 4.157127031019202e-05,
      "loss": 0.0169,
      "step": 963
    },
    {
      "epoch": 0.35,
      "learning_rate": 4.156203840472674e-05,
      "loss": 0.0566,
      "step": 964
    },
    {
      "epoch": 0.35,
      "learning_rate": 4.155280649926145e-05,
      "loss": 0.1859,
      "step": 965
    },
    {
      "epoch": 0.35,
      "learning_rate": 4.1543574593796166e-05,
      "loss": 0.2344,
      "step": 966
    },
    {
      "epoch": 0.35,
      "learning_rate": 4.153434268833088e-05,
      "loss": 0.02,
      "step": 967
    },
    {
      "epoch": 0.35,
      "learning_rate": 4.152511078286558e-05,
      "loss": 0.0116,
      "step": 968
    },
    {
      "epoch": 0.35,
      "learning_rate": 4.15158788774003e-05,
      "loss": 0.2994,
      "step": 969
    },
    {
      "epoch": 0.35,
      "learning_rate": 4.150664697193501e-05,
      "loss": 0.0241,
      "step": 970
    },
    {
      "epoch": 0.36,
      "learning_rate": 4.149741506646972e-05,
      "loss": 0.0097,
      "step": 971
    },
    {
      "epoch": 0.36,
      "learning_rate": 4.1488183161004436e-05,
      "loss": 0.4325,
      "step": 972
    },
    {
      "epoch": 0.36,
      "learning_rate": 4.147895125553915e-05,
      "loss": 0.2331,
      "step": 973
    },
    {
      "epoch": 0.36,
      "learning_rate": 4.146971935007386e-05,
      "loss": 0.0154,
      "step": 974
    },
    {
      "epoch": 0.36,
      "learning_rate": 4.146048744460857e-05,
      "loss": 0.2686,
      "step": 975
    },
    {
      "epoch": 0.36,
      "learning_rate": 4.145125553914328e-05,
      "loss": 0.0267,
      "step": 976
    },
    {
      "epoch": 0.36,
      "learning_rate": 4.1442023633677996e-05,
      "loss": 0.6117,
      "step": 977
    },
    {
      "epoch": 0.36,
      "learning_rate": 4.1432791728212707e-05,
      "loss": 0.1714,
      "step": 978
    },
    {
      "epoch": 0.36,
      "learning_rate": 4.142355982274742e-05,
      "loss": 0.0402,
      "step": 979
    },
    {
      "epoch": 0.36,
      "learning_rate": 4.141432791728213e-05,
      "loss": 0.1897,
      "step": 980
    },
    {
      "epoch": 0.36,
      "learning_rate": 4.140509601181684e-05,
      "loss": 0.4051,
      "step": 981
    },
    {
      "epoch": 0.36,
      "learning_rate": 4.139586410635155e-05,
      "loss": 0.0072,
      "step": 982
    },
    {
      "epoch": 0.36,
      "learning_rate": 4.1386632200886266e-05,
      "loss": 0.5042,
      "step": 983
    },
    {
      "epoch": 0.36,
      "learning_rate": 4.1377400295420977e-05,
      "loss": 0.9264,
      "step": 984
    },
    {
      "epoch": 0.36,
      "learning_rate": 4.1368168389955694e-05,
      "loss": 0.4266,
      "step": 985
    },
    {
      "epoch": 0.36,
      "learning_rate": 4.1358936484490404e-05,
      "loss": 0.4766,
      "step": 986
    },
    {
      "epoch": 0.36,
      "learning_rate": 4.134970457902511e-05,
      "loss": 0.2206,
      "step": 987
    },
    {
      "epoch": 0.36,
      "learning_rate": 4.1340472673559826e-05,
      "loss": 0.1039,
      "step": 988
    },
    {
      "epoch": 0.36,
      "learning_rate": 4.1331240768094536e-05,
      "loss": 0.1934,
      "step": 989
    },
    {
      "epoch": 0.36,
      "learning_rate": 4.132200886262925e-05,
      "loss": 0.0637,
      "step": 990
    },
    {
      "epoch": 0.36,
      "learning_rate": 4.1312776957163964e-05,
      "loss": 0.1198,
      "step": 991
    },
    {
      "epoch": 0.36,
      "learning_rate": 4.1303545051698675e-05,
      "loss": 0.0287,
      "step": 992
    },
    {
      "epoch": 0.36,
      "learning_rate": 4.1294313146233385e-05,
      "loss": 0.1636,
      "step": 993
    },
    {
      "epoch": 0.36,
      "learning_rate": 4.1285081240768096e-05,
      "loss": 0.0348,
      "step": 994
    },
    {
      "epoch": 0.36,
      "learning_rate": 4.1275849335302806e-05,
      "loss": 0.2227,
      "step": 995
    },
    {
      "epoch": 0.36,
      "learning_rate": 4.1266617429837523e-05,
      "loss": 0.1658,
      "step": 996
    },
    {
      "epoch": 0.36,
      "learning_rate": 4.1257385524372234e-05,
      "loss": 0.2154,
      "step": 997
    },
    {
      "epoch": 0.37,
      "learning_rate": 4.1248153618906945e-05,
      "loss": 0.0419,
      "step": 998
    },
    {
      "epoch": 0.37,
      "learning_rate": 4.1238921713441655e-05,
      "loss": 0.3221,
      "step": 999
    },
    {
      "epoch": 0.37,
      "learning_rate": 4.1229689807976366e-05,
      "loss": 0.0179,
      "step": 1000
    },
    {
      "epoch": 0.37,
      "learning_rate": 4.1220457902511076e-05,
      "loss": 0.2962,
      "step": 1001
    },
    {
      "epoch": 0.37,
      "learning_rate": 4.1211225997045794e-05,
      "loss": 0.0274,
      "step": 1002
    },
    {
      "epoch": 0.37,
      "learning_rate": 4.1201994091580504e-05,
      "loss": 0.1111,
      "step": 1003
    },
    {
      "epoch": 0.37,
      "learning_rate": 4.1192762186115215e-05,
      "loss": 0.0451,
      "step": 1004
    },
    {
      "epoch": 0.37,
      "learning_rate": 4.118353028064993e-05,
      "loss": 0.0452,
      "step": 1005
    },
    {
      "epoch": 0.37,
      "learning_rate": 4.1174298375184636e-05,
      "loss": 0.201,
      "step": 1006
    },
    {
      "epoch": 0.37,
      "learning_rate": 4.116506646971935e-05,
      "loss": 0.2196,
      "step": 1007
    },
    {
      "epoch": 0.37,
      "learning_rate": 4.1155834564254064e-05,
      "loss": 0.2675,
      "step": 1008
    },
    {
      "epoch": 0.37,
      "learning_rate": 4.1146602658788774e-05,
      "loss": 0.0939,
      "step": 1009
    },
    {
      "epoch": 0.37,
      "learning_rate": 4.113737075332349e-05,
      "loss": 0.075,
      "step": 1010
    },
    {
      "epoch": 0.37,
      "learning_rate": 4.11281388478582e-05,
      "loss": 0.1244,
      "step": 1011
    },
    {
      "epoch": 0.37,
      "learning_rate": 4.111890694239291e-05,
      "loss": 0.0706,
      "step": 1012
    },
    {
      "epoch": 0.37,
      "learning_rate": 4.110967503692762e-05,
      "loss": 0.0826,
      "step": 1013
    },
    {
      "epoch": 0.37,
      "learning_rate": 4.1100443131462334e-05,
      "loss": 0.0199,
      "step": 1014
    },
    {
      "epoch": 0.37,
      "learning_rate": 4.109121122599705e-05,
      "loss": 0.5214,
      "step": 1015
    },
    {
      "epoch": 0.37,
      "learning_rate": 4.108197932053176e-05,
      "loss": 0.0411,
      "step": 1016
    },
    {
      "epoch": 0.37,
      "learning_rate": 4.107274741506647e-05,
      "loss": 0.0228,
      "step": 1017
    },
    {
      "epoch": 0.37,
      "learning_rate": 4.106351550960118e-05,
      "loss": 0.2236,
      "step": 1018
    },
    {
      "epoch": 0.37,
      "learning_rate": 4.105428360413589e-05,
      "loss": 0.1476,
      "step": 1019
    },
    {
      "epoch": 0.37,
      "learning_rate": 4.1045051698670604e-05,
      "loss": 0.025,
      "step": 1020
    },
    {
      "epoch": 0.37,
      "learning_rate": 4.103581979320532e-05,
      "loss": 0.0331,
      "step": 1021
    },
    {
      "epoch": 0.37,
      "learning_rate": 4.102658788774003e-05,
      "loss": 0.196,
      "step": 1022
    },
    {
      "epoch": 0.37,
      "learning_rate": 4.101735598227474e-05,
      "loss": 0.0747,
      "step": 1023
    },
    {
      "epoch": 0.37,
      "learning_rate": 4.100812407680946e-05,
      "loss": 0.0595,
      "step": 1024
    },
    {
      "epoch": 0.38,
      "learning_rate": 4.099889217134416e-05,
      "loss": 0.2338,
      "step": 1025
    },
    {
      "epoch": 0.38,
      "learning_rate": 4.098966026587888e-05,
      "loss": 0.0384,
      "step": 1026
    },
    {
      "epoch": 0.38,
      "learning_rate": 4.098042836041359e-05,
      "loss": 0.3588,
      "step": 1027
    },
    {
      "epoch": 0.38,
      "learning_rate": 4.09711964549483e-05,
      "loss": 0.1988,
      "step": 1028
    },
    {
      "epoch": 0.38,
      "learning_rate": 4.096196454948302e-05,
      "loss": 0.1046,
      "step": 1029
    },
    {
      "epoch": 0.38,
      "learning_rate": 4.095273264401773e-05,
      "loss": 0.548,
      "step": 1030
    },
    {
      "epoch": 0.38,
      "learning_rate": 4.094350073855244e-05,
      "loss": 0.2818,
      "step": 1031
    },
    {
      "epoch": 0.38,
      "learning_rate": 4.093426883308715e-05,
      "loss": 0.0261,
      "step": 1032
    },
    {
      "epoch": 0.38,
      "learning_rate": 4.092503692762186e-05,
      "loss": 0.0136,
      "step": 1033
    },
    {
      "epoch": 0.38,
      "learning_rate": 4.091580502215657e-05,
      "loss": 0.0113,
      "step": 1034
    },
    {
      "epoch": 0.38,
      "learning_rate": 4.090657311669129e-05,
      "loss": 0.1439,
      "step": 1035
    },
    {
      "epoch": 0.38,
      "learning_rate": 4.0897341211226e-05,
      "loss": 0.0414,
      "step": 1036
    },
    {
      "epoch": 0.38,
      "learning_rate": 4.088810930576071e-05,
      "loss": 0.174,
      "step": 1037
    },
    {
      "epoch": 0.38,
      "learning_rate": 4.087887740029542e-05,
      "loss": 0.0319,
      "step": 1038
    },
    {
      "epoch": 0.38,
      "learning_rate": 4.086964549483013e-05,
      "loss": 0.0094,
      "step": 1039
    },
    {
      "epoch": 0.38,
      "learning_rate": 4.086041358936485e-05,
      "loss": 0.0145,
      "step": 1040
    },
    {
      "epoch": 0.38,
      "learning_rate": 4.085118168389956e-05,
      "loss": 0.187,
      "step": 1041
    },
    {
      "epoch": 0.38,
      "learning_rate": 4.084194977843427e-05,
      "loss": 0.1926,
      "step": 1042
    },
    {
      "epoch": 0.38,
      "learning_rate": 4.083271787296899e-05,
      "loss": 0.0095,
      "step": 1043
    },
    {
      "epoch": 0.38,
      "learning_rate": 4.082348596750369e-05,
      "loss": 0.3351,
      "step": 1044
    },
    {
      "epoch": 0.38,
      "learning_rate": 4.08142540620384e-05,
      "loss": 0.0076,
      "step": 1045
    },
    {
      "epoch": 0.38,
      "learning_rate": 4.080502215657312e-05,
      "loss": 0.0237,
      "step": 1046
    },
    {
      "epoch": 0.38,
      "learning_rate": 4.079579025110783e-05,
      "loss": 0.0333,
      "step": 1047
    },
    {
      "epoch": 0.38,
      "learning_rate": 4.0786558345642547e-05,
      "loss": 0.1115,
      "step": 1048
    },
    {
      "epoch": 0.38,
      "learning_rate": 4.077732644017726e-05,
      "loss": 0.5512,
      "step": 1049
    },
    {
      "epoch": 0.38,
      "learning_rate": 4.076809453471197e-05,
      "loss": 0.0125,
      "step": 1050
    },
    {
      "epoch": 0.38,
      "learning_rate": 4.075886262924668e-05,
      "loss": 0.5607,
      "step": 1051
    },
    {
      "epoch": 0.38,
      "learning_rate": 4.074963072378139e-05,
      "loss": 0.7918,
      "step": 1052
    },
    {
      "epoch": 0.39,
      "learning_rate": 4.07403988183161e-05,
      "loss": 0.0306,
      "step": 1053
    },
    {
      "epoch": 0.39,
      "learning_rate": 4.073116691285082e-05,
      "loss": 0.0315,
      "step": 1054
    },
    {
      "epoch": 0.39,
      "learning_rate": 4.072193500738553e-05,
      "loss": 0.0088,
      "step": 1055
    },
    {
      "epoch": 0.39,
      "learning_rate": 4.071270310192024e-05,
      "loss": 0.2773,
      "step": 1056
    },
    {
      "epoch": 0.39,
      "learning_rate": 4.070347119645495e-05,
      "loss": 0.2322,
      "step": 1057
    },
    {
      "epoch": 0.39,
      "learning_rate": 4.069423929098966e-05,
      "loss": 0.0083,
      "step": 1058
    },
    {
      "epoch": 0.39,
      "learning_rate": 4.0685007385524376e-05,
      "loss": 0.3493,
      "step": 1059
    },
    {
      "epoch": 0.39,
      "learning_rate": 4.067577548005909e-05,
      "loss": 0.2386,
      "step": 1060
    },
    {
      "epoch": 0.39,
      "learning_rate": 4.06665435745938e-05,
      "loss": 0.0383,
      "step": 1061
    },
    {
      "epoch": 0.39,
      "learning_rate": 4.0657311669128515e-05,
      "loss": 0.4568,
      "step": 1062
    },
    {
      "epoch": 0.39,
      "learning_rate": 4.064807976366322e-05,
      "loss": 0.0119,
      "step": 1063
    },
    {
      "epoch": 0.39,
      "learning_rate": 4.063884785819793e-05,
      "loss": 0.7905,
      "step": 1064
    },
    {
      "epoch": 0.39,
      "learning_rate": 4.0629615952732646e-05,
      "loss": 0.0433,
      "step": 1065
    },
    {
      "epoch": 0.39,
      "learning_rate": 4.062038404726736e-05,
      "loss": 0.0073,
      "step": 1066
    },
    {
      "epoch": 0.39,
      "learning_rate": 4.0611152141802074e-05,
      "loss": 0.0769,
      "step": 1067
    },
    {
      "epoch": 0.39,
      "learning_rate": 4.0601920236336785e-05,
      "loss": 0.2324,
      "step": 1068
    },
    {
      "epoch": 0.39,
      "learning_rate": 4.0592688330871495e-05,
      "loss": 0.2667,
      "step": 1069
    },
    {
      "epoch": 0.39,
      "learning_rate": 4.0583456425406206e-05,
      "loss": 0.1118,
      "step": 1070
    },
    {
      "epoch": 0.39,
      "learning_rate": 4.0574224519940916e-05,
      "loss": 0.0875,
      "step": 1071
    },
    {
      "epoch": 0.39,
      "learning_rate": 4.056499261447563e-05,
      "loss": 0.0446,
      "step": 1072
    },
    {
      "epoch": 0.39,
      "learning_rate": 4.0555760709010344e-05,
      "loss": 0.0121,
      "step": 1073
    },
    {
      "epoch": 0.39,
      "learning_rate": 4.0546528803545055e-05,
      "loss": 0.3899,
      "step": 1074
    },
    {
      "epoch": 0.39,
      "learning_rate": 4.0537296898079765e-05,
      "loss": 0.3342,
      "step": 1075
    },
    {
      "epoch": 0.39,
      "learning_rate": 4.0528064992614476e-05,
      "loss": 0.0495,
      "step": 1076
    },
    {
      "epoch": 0.39,
      "learning_rate": 4.0518833087149186e-05,
      "loss": 0.1889,
      "step": 1077
    },
    {
      "epoch": 0.39,
      "learning_rate": 4.0509601181683904e-05,
      "loss": 0.0501,
      "step": 1078
    },
    {
      "epoch": 0.39,
      "learning_rate": 4.0500369276218614e-05,
      "loss": 0.1306,
      "step": 1079
    },
    {
      "epoch": 0.4,
      "learning_rate": 4.0491137370753325e-05,
      "loss": 0.2318,
      "step": 1080
    },
    {
      "epoch": 0.4,
      "learning_rate": 4.048190546528804e-05,
      "loss": 0.3667,
      "step": 1081
    },
    {
      "epoch": 0.4,
      "learning_rate": 4.0472673559822746e-05,
      "loss": 0.2479,
      "step": 1082
    },
    {
      "epoch": 0.4,
      "learning_rate": 4.0463441654357456e-05,
      "loss": 0.0229,
      "step": 1083
    },
    {
      "epoch": 0.4,
      "learning_rate": 4.0454209748892174e-05,
      "loss": 0.4013,
      "step": 1084
    },
    {
      "epoch": 0.4,
      "learning_rate": 4.0444977843426884e-05,
      "loss": 0.0529,
      "step": 1085
    },
    {
      "epoch": 0.4,
      "learning_rate": 4.04357459379616e-05,
      "loss": 0.3128,
      "step": 1086
    },
    {
      "epoch": 0.4,
      "learning_rate": 4.042651403249631e-05,
      "loss": 0.1823,
      "step": 1087
    },
    {
      "epoch": 0.4,
      "learning_rate": 4.041728212703102e-05,
      "loss": 0.0745,
      "step": 1088
    },
    {
      "epoch": 0.4,
      "learning_rate": 4.040805022156573e-05,
      "loss": 0.1824,
      "step": 1089
    },
    {
      "epoch": 0.4,
      "learning_rate": 4.0398818316100444e-05,
      "loss": 0.079,
      "step": 1090
    },
    {
      "epoch": 0.4,
      "learning_rate": 4.0389586410635154e-05,
      "loss": 0.3556,
      "step": 1091
    },
    {
      "epoch": 0.4,
      "learning_rate": 4.038035450516987e-05,
      "loss": 0.1839,
      "step": 1092
    },
    {
      "epoch": 0.4,
      "learning_rate": 4.037112259970458e-05,
      "loss": 0.2248,
      "step": 1093
    },
    {
      "epoch": 0.4,
      "learning_rate": 4.036189069423929e-05,
      "loss": 0.1494,
      "step": 1094
    },
    {
      "epoch": 0.4,
      "learning_rate": 4.0352658788774e-05,
      "loss": 0.0474,
      "step": 1095
    },
    {
      "epoch": 0.4,
      "learning_rate": 4.0343426883308714e-05,
      "loss": 0.0436,
      "step": 1096
    },
    {
      "epoch": 0.4,
      "learning_rate": 4.033419497784343e-05,
      "loss": 0.0387,
      "step": 1097
    },
    {
      "epoch": 0.4,
      "learning_rate": 4.032496307237814e-05,
      "loss": 0.0208,
      "step": 1098
    },
    {
      "epoch": 0.4,
      "learning_rate": 4.031573116691285e-05,
      "loss": 0.0332,
      "step": 1099
    },
    {
      "epoch": 0.4,
      "learning_rate": 4.030649926144757e-05,
      "loss": 0.0401,
      "step": 1100
    },
    {
      "epoch": 0.4,
      "learning_rate": 4.0297267355982273e-05,
      "loss": 0.043,
      "step": 1101
    },
    {
      "epoch": 0.4,
      "learning_rate": 4.0288035450516984e-05,
      "loss": 0.1405,
      "step": 1102
    },
    {
      "epoch": 0.4,
      "learning_rate": 4.02788035450517e-05,
      "loss": 0.1407,
      "step": 1103
    },
    {
      "epoch": 0.4,
      "learning_rate": 4.026957163958641e-05,
      "loss": 0.0143,
      "step": 1104
    },
    {
      "epoch": 0.4,
      "learning_rate": 4.026033973412112e-05,
      "loss": 0.0079,
      "step": 1105
    },
    {
      "epoch": 0.4,
      "learning_rate": 4.025110782865584e-05,
      "loss": 0.0143,
      "step": 1106
    },
    {
      "epoch": 0.41,
      "learning_rate": 4.024187592319055e-05,
      "loss": 0.2704,
      "step": 1107
    },
    {
      "epoch": 0.41,
      "learning_rate": 4.023264401772526e-05,
      "loss": 0.1412,
      "step": 1108
    },
    {
      "epoch": 0.41,
      "learning_rate": 4.022341211225997e-05,
      "loss": 0.0049,
      "step": 1109
    },
    {
      "epoch": 0.41,
      "learning_rate": 4.021418020679468e-05,
      "loss": 0.1533,
      "step": 1110
    },
    {
      "epoch": 0.41,
      "learning_rate": 4.02049483013294e-05,
      "loss": 0.0058,
      "step": 1111
    },
    {
      "epoch": 0.41,
      "learning_rate": 4.019571639586411e-05,
      "loss": 0.197,
      "step": 1112
    },
    {
      "epoch": 0.41,
      "learning_rate": 4.018648449039882e-05,
      "loss": 0.0799,
      "step": 1113
    },
    {
      "epoch": 0.41,
      "learning_rate": 4.017725258493353e-05,
      "loss": 0.0027,
      "step": 1114
    },
    {
      "epoch": 0.41,
      "learning_rate": 4.016802067946824e-05,
      "loss": 0.3467,
      "step": 1115
    },
    {
      "epoch": 0.41,
      "learning_rate": 4.015878877400295e-05,
      "loss": 0.0141,
      "step": 1116
    },
    {
      "epoch": 0.41,
      "learning_rate": 4.014955686853767e-05,
      "loss": 0.0041,
      "step": 1117
    },
    {
      "epoch": 0.41,
      "learning_rate": 4.014032496307238e-05,
      "loss": 0.0034,
      "step": 1118
    },
    {
      "epoch": 0.41,
      "learning_rate": 4.01310930576071e-05,
      "loss": 0.6238,
      "step": 1119
    },
    {
      "epoch": 0.41,
      "learning_rate": 4.01218611521418e-05,
      "loss": 0.0314,
      "step": 1120
    },
    {
      "epoch": 0.41,
      "learning_rate": 4.011262924667651e-05,
      "loss": 0.008,
      "step": 1121
    },
    {
      "epoch": 0.41,
      "learning_rate": 4.010339734121123e-05,
      "loss": 0.0024,
      "step": 1122
    },
    {
      "epoch": 0.41,
      "learning_rate": 4.009416543574594e-05,
      "loss": 0.5156,
      "step": 1123
    },
    {
      "epoch": 0.41,
      "learning_rate": 4.008493353028065e-05,
      "loss": 0.0079,
      "step": 1124
    },
    {
      "epoch": 0.41,
      "learning_rate": 4.007570162481537e-05,
      "loss": 0.3497,
      "step": 1125
    },
    {
      "epoch": 0.41,
      "learning_rate": 4.006646971935008e-05,
      "loss": 0.0021,
      "step": 1126
    },
    {
      "epoch": 0.41,
      "learning_rate": 4.005723781388479e-05,
      "loss": 0.0049,
      "step": 1127
    },
    {
      "epoch": 0.41,
      "learning_rate": 4.00480059084195e-05,
      "loss": 0.0171,
      "step": 1128
    },
    {
      "epoch": 0.41,
      "learning_rate": 4.003877400295421e-05,
      "loss": 0.6202,
      "step": 1129
    },
    {
      "epoch": 0.41,
      "learning_rate": 4.002954209748893e-05,
      "loss": 0.6042,
      "step": 1130
    },
    {
      "epoch": 0.41,
      "learning_rate": 4.002031019202364e-05,
      "loss": 0.0108,
      "step": 1131
    },
    {
      "epoch": 0.41,
      "learning_rate": 4.001107828655835e-05,
      "loss": 0.2527,
      "step": 1132
    },
    {
      "epoch": 0.41,
      "learning_rate": 4.000184638109306e-05,
      "loss": 0.0187,
      "step": 1133
    },
    {
      "epoch": 0.41,
      "learning_rate": 3.999261447562777e-05,
      "loss": 0.4862,
      "step": 1134
    },
    {
      "epoch": 0.42,
      "learning_rate": 3.998338257016248e-05,
      "loss": 0.3067,
      "step": 1135
    },
    {
      "epoch": 0.42,
      "learning_rate": 3.99741506646972e-05,
      "loss": 0.0053,
      "step": 1136
    },
    {
      "epoch": 0.42,
      "learning_rate": 3.996491875923191e-05,
      "loss": 0.0105,
      "step": 1137
    },
    {
      "epoch": 0.42,
      "learning_rate": 3.9955686853766625e-05,
      "loss": 0.1986,
      "step": 1138
    },
    {
      "epoch": 0.42,
      "learning_rate": 3.994645494830133e-05,
      "loss": 0.0227,
      "step": 1139
    },
    {
      "epoch": 0.42,
      "learning_rate": 3.993722304283604e-05,
      "loss": 0.0126,
      "step": 1140
    },
    {
      "epoch": 0.42,
      "learning_rate": 3.9927991137370756e-05,
      "loss": 0.2248,
      "step": 1141
    },
    {
      "epoch": 0.42,
      "learning_rate": 3.991875923190547e-05,
      "loss": 0.3419,
      "step": 1142
    },
    {
      "epoch": 0.42,
      "learning_rate": 3.990952732644018e-05,
      "loss": 0.023,
      "step": 1143
    },
    {
      "epoch": 0.42,
      "learning_rate": 3.9900295420974895e-05,
      "loss": 0.0999,
      "step": 1144
    },
    {
      "epoch": 0.42,
      "learning_rate": 3.9891063515509605e-05,
      "loss": 0.02,
      "step": 1145
    },
    {
      "epoch": 0.42,
      "learning_rate": 3.988183161004431e-05,
      "loss": 0.2081,
      "step": 1146
    },
    {
      "epoch": 0.42,
      "learning_rate": 3.9872599704579026e-05,
      "loss": 0.5302,
      "step": 1147
    },
    {
      "epoch": 0.42,
      "learning_rate": 3.986336779911374e-05,
      "loss": 0.0931,
      "step": 1148
    },
    {
      "epoch": 0.42,
      "learning_rate": 3.9854135893648454e-05,
      "loss": 0.0117,
      "step": 1149
    },
    {
      "epoch": 0.42,
      "learning_rate": 3.9844903988183165e-05,
      "loss": 0.0316,
      "step": 1150
    },
    {
      "epoch": 0.42,
      "learning_rate": 3.9835672082717875e-05,
      "loss": 0.1474,
      "step": 1151
    },
    {
      "epoch": 0.42,
      "learning_rate": 3.9826440177252586e-05,
      "loss": 0.0103,
      "step": 1152
    },
    {
      "epoch": 0.42,
      "learning_rate": 3.9817208271787297e-05,
      "loss": 0.0151,
      "step": 1153
    },
    {
      "epoch": 0.42,
      "learning_rate": 3.980797636632201e-05,
      "loss": 0.0152,
      "step": 1154
    },
    {
      "epoch": 0.42,
      "learning_rate": 3.9798744460856724e-05,
      "loss": 0.2848,
      "step": 1155
    },
    {
      "epoch": 0.42,
      "learning_rate": 3.9789512555391435e-05,
      "loss": 0.0236,
      "step": 1156
    },
    {
      "epoch": 0.42,
      "learning_rate": 3.978028064992615e-05,
      "loss": 0.0093,
      "step": 1157
    },
    {
      "epoch": 0.42,
      "learning_rate": 3.9771048744460856e-05,
      "loss": 0.1224,
      "step": 1158
    },
    {
      "epoch": 0.42,
      "learning_rate": 3.9761816838995567e-05,
      "loss": 0.0062,
      "step": 1159
    },
    {
      "epoch": 0.42,
      "learning_rate": 3.9752584933530284e-05,
      "loss": 0.0043,
      "step": 1160
    },
    {
      "epoch": 0.42,
      "learning_rate": 3.9743353028064994e-05,
      "loss": 0.2867,
      "step": 1161
    },
    {
      "epoch": 0.43,
      "learning_rate": 3.9734121122599705e-05,
      "loss": 0.5611,
      "step": 1162
    },
    {
      "epoch": 0.43,
      "learning_rate": 3.972488921713442e-05,
      "loss": 0.1897,
      "step": 1163
    },
    {
      "epoch": 0.43,
      "learning_rate": 3.971565731166913e-05,
      "loss": 0.5786,
      "step": 1164
    },
    {
      "epoch": 0.43,
      "learning_rate": 3.970642540620384e-05,
      "loss": 0.6381,
      "step": 1165
    },
    {
      "epoch": 0.43,
      "learning_rate": 3.9697193500738554e-05,
      "loss": 0.01,
      "step": 1166
    },
    {
      "epoch": 0.43,
      "learning_rate": 3.9687961595273265e-05,
      "loss": 0.005,
      "step": 1167
    },
    {
      "epoch": 0.43,
      "learning_rate": 3.967872968980798e-05,
      "loss": 0.0067,
      "step": 1168
    },
    {
      "epoch": 0.43,
      "learning_rate": 3.966949778434269e-05,
      "loss": 0.0153,
      "step": 1169
    },
    {
      "epoch": 0.43,
      "learning_rate": 3.96602658788774e-05,
      "loss": 0.0156,
      "step": 1170
    },
    {
      "epoch": 0.43,
      "learning_rate": 3.9651033973412114e-05,
      "loss": 0.5901,
      "step": 1171
    },
    {
      "epoch": 0.43,
      "learning_rate": 3.9641802067946824e-05,
      "loss": 0.01,
      "step": 1172
    },
    {
      "epoch": 0.43,
      "learning_rate": 3.9632570162481535e-05,
      "loss": 0.0498,
      "step": 1173
    },
    {
      "epoch": 0.43,
      "learning_rate": 3.962333825701625e-05,
      "loss": 0.0103,
      "step": 1174
    },
    {
      "epoch": 0.43,
      "learning_rate": 3.961410635155096e-05,
      "loss": 0.3355,
      "step": 1175
    },
    {
      "epoch": 0.43,
      "learning_rate": 3.960487444608567e-05,
      "loss": 0.0127,
      "step": 1176
    },
    {
      "epoch": 0.43,
      "learning_rate": 3.9595642540620384e-05,
      "loss": 0.0084,
      "step": 1177
    },
    {
      "epoch": 0.43,
      "learning_rate": 3.9586410635155094e-05,
      "loss": 0.0129,
      "step": 1178
    },
    {
      "epoch": 0.43,
      "learning_rate": 3.957717872968981e-05,
      "loss": 0.4899,
      "step": 1179
    },
    {
      "epoch": 0.43,
      "learning_rate": 3.956794682422452e-05,
      "loss": 0.3853,
      "step": 1180
    },
    {
      "epoch": 0.43,
      "learning_rate": 3.955871491875923e-05,
      "loss": 0.2701,
      "step": 1181
    },
    {
      "epoch": 0.43,
      "learning_rate": 3.954948301329395e-05,
      "loss": 0.1037,
      "step": 1182
    },
    {
      "epoch": 0.43,
      "learning_rate": 3.954025110782866e-05,
      "loss": 0.0446,
      "step": 1183
    },
    {
      "epoch": 0.43,
      "learning_rate": 3.9531019202363364e-05,
      "loss": 0.0104,
      "step": 1184
    },
    {
      "epoch": 0.43,
      "learning_rate": 3.952178729689808e-05,
      "loss": 0.4461,
      "step": 1185
    },
    {
      "epoch": 0.43,
      "learning_rate": 3.951255539143279e-05,
      "loss": 0.0271,
      "step": 1186
    },
    {
      "epoch": 0.43,
      "learning_rate": 3.950332348596751e-05,
      "loss": 0.0078,
      "step": 1187
    },
    {
      "epoch": 0.43,
      "learning_rate": 3.949409158050222e-05,
      "loss": 0.0167,
      "step": 1188
    },
    {
      "epoch": 0.44,
      "learning_rate": 3.948485967503693e-05,
      "loss": 0.0131,
      "step": 1189
    },
    {
      "epoch": 0.44,
      "learning_rate": 3.947562776957164e-05,
      "loss": 0.0201,
      "step": 1190
    },
    {
      "epoch": 0.44,
      "learning_rate": 3.946639586410635e-05,
      "loss": 0.2075,
      "step": 1191
    },
    {
      "epoch": 0.44,
      "learning_rate": 3.945716395864106e-05,
      "loss": 0.0275,
      "step": 1192
    },
    {
      "epoch": 0.44,
      "learning_rate": 3.944793205317578e-05,
      "loss": 0.0604,
      "step": 1193
    },
    {
      "epoch": 0.44,
      "learning_rate": 3.943870014771049e-05,
      "loss": 0.0456,
      "step": 1194
    },
    {
      "epoch": 0.44,
      "learning_rate": 3.94294682422452e-05,
      "loss": 0.0131,
      "step": 1195
    },
    {
      "epoch": 0.44,
      "learning_rate": 3.942023633677991e-05,
      "loss": 0.0049,
      "step": 1196
    },
    {
      "epoch": 0.44,
      "learning_rate": 3.941100443131462e-05,
      "loss": 0.0089,
      "step": 1197
    },
    {
      "epoch": 0.44,
      "learning_rate": 3.940177252584934e-05,
      "loss": 0.1892,
      "step": 1198
    },
    {
      "epoch": 0.44,
      "learning_rate": 3.939254062038405e-05,
      "loss": 0.0123,
      "step": 1199
    },
    {
      "epoch": 0.44,
      "learning_rate": 3.938330871491876e-05,
      "loss": 0.0053,
      "step": 1200
    },
    {
      "epoch": 0.44,
      "learning_rate": 3.937407680945348e-05,
      "loss": 0.1838,
      "step": 1201
    },
    {
      "epoch": 0.44,
      "learning_rate": 3.936484490398819e-05,
      "loss": 0.0293,
      "step": 1202
    },
    {
      "epoch": 0.44,
      "learning_rate": 3.935561299852289e-05,
      "loss": 0.0042,
      "step": 1203
    },
    {
      "epoch": 0.44,
      "learning_rate": 3.934638109305761e-05,
      "loss": 0.2056,
      "step": 1204
    },
    {
      "epoch": 0.44,
      "learning_rate": 3.933714918759232e-05,
      "loss": 0.485,
      "step": 1205
    },
    {
      "epoch": 0.44,
      "learning_rate": 3.932791728212703e-05,
      "loss": 0.3997,
      "step": 1206
    },
    {
      "epoch": 0.44,
      "learning_rate": 3.931868537666175e-05,
      "loss": 0.1322,
      "step": 1207
    },
    {
      "epoch": 0.44,
      "learning_rate": 3.930945347119646e-05,
      "loss": 0.3175,
      "step": 1208
    },
    {
      "epoch": 0.44,
      "learning_rate": 3.930022156573117e-05,
      "loss": 0.1557,
      "step": 1209
    },
    {
      "epoch": 0.44,
      "learning_rate": 3.929098966026588e-05,
      "loss": 0.2888,
      "step": 1210
    },
    {
      "epoch": 0.44,
      "learning_rate": 3.928175775480059e-05,
      "loss": 0.1068,
      "step": 1211
    },
    {
      "epoch": 0.44,
      "learning_rate": 3.927252584933531e-05,
      "loss": 0.0071,
      "step": 1212
    },
    {
      "epoch": 0.44,
      "learning_rate": 3.926329394387002e-05,
      "loss": 1.1313,
      "step": 1213
    },
    {
      "epoch": 0.44,
      "learning_rate": 3.925406203840473e-05,
      "loss": 0.2301,
      "step": 1214
    },
    {
      "epoch": 0.44,
      "learning_rate": 3.924483013293944e-05,
      "loss": 0.2467,
      "step": 1215
    },
    {
      "epoch": 0.44,
      "learning_rate": 3.923559822747415e-05,
      "loss": 0.2241,
      "step": 1216
    },
    {
      "epoch": 0.45,
      "learning_rate": 3.922636632200886e-05,
      "loss": 0.3473,
      "step": 1217
    },
    {
      "epoch": 0.45,
      "learning_rate": 3.921713441654358e-05,
      "loss": 0.0302,
      "step": 1218
    },
    {
      "epoch": 0.45,
      "learning_rate": 3.920790251107829e-05,
      "loss": 0.4725,
      "step": 1219
    },
    {
      "epoch": 0.45,
      "learning_rate": 3.9198670605613005e-05,
      "loss": 0.2412,
      "step": 1220
    },
    {
      "epoch": 0.45,
      "learning_rate": 3.9189438700147716e-05,
      "loss": 0.3422,
      "step": 1221
    },
    {
      "epoch": 0.45,
      "learning_rate": 3.918020679468242e-05,
      "loss": 0.3743,
      "step": 1222
    },
    {
      "epoch": 0.45,
      "learning_rate": 3.917097488921714e-05,
      "loss": 0.008,
      "step": 1223
    },
    {
      "epoch": 0.45,
      "learning_rate": 3.916174298375185e-05,
      "loss": 0.0785,
      "step": 1224
    },
    {
      "epoch": 0.45,
      "learning_rate": 3.915251107828656e-05,
      "loss": 0.0637,
      "step": 1225
    },
    {
      "epoch": 0.45,
      "learning_rate": 3.9143279172821275e-05,
      "loss": 0.2007,
      "step": 1226
    },
    {
      "epoch": 0.45,
      "learning_rate": 3.9134047267355986e-05,
      "loss": 0.0276,
      "step": 1227
    },
    {
      "epoch": 0.45,
      "learning_rate": 3.9124815361890696e-05,
      "loss": 0.1257,
      "step": 1228
    },
    {
      "epoch": 0.45,
      "learning_rate": 3.911558345642541e-05,
      "loss": 0.0959,
      "step": 1229
    },
    {
      "epoch": 0.45,
      "learning_rate": 3.910635155096012e-05,
      "loss": 0.1972,
      "step": 1230
    },
    {
      "epoch": 0.45,
      "learning_rate": 3.9097119645494835e-05,
      "loss": 0.4506,
      "step": 1231
    },
    {
      "epoch": 0.45,
      "learning_rate": 3.9087887740029545e-05,
      "loss": 1.1447,
      "step": 1232
    },
    {
      "epoch": 0.45,
      "learning_rate": 3.9078655834564256e-05,
      "loss": 0.933,
      "step": 1233
    },
    {
      "epoch": 0.45,
      "learning_rate": 3.9069423929098966e-05,
      "loss": 0.0564,
      "step": 1234
    },
    {
      "epoch": 0.45,
      "learning_rate": 3.906019202363368e-05,
      "loss": 0.0213,
      "step": 1235
    },
    {
      "epoch": 0.45,
      "learning_rate": 3.905096011816839e-05,
      "loss": 0.3461,
      "step": 1236
    },
    {
      "epoch": 0.45,
      "learning_rate": 3.9041728212703105e-05,
      "loss": 0.2755,
      "step": 1237
    },
    {
      "epoch": 0.45,
      "learning_rate": 3.9032496307237815e-05,
      "loss": 0.0766,
      "step": 1238
    },
    {
      "epoch": 0.45,
      "learning_rate": 3.902326440177253e-05,
      "loss": 0.335,
      "step": 1239
    },
    {
      "epoch": 0.45,
      "learning_rate": 3.901403249630724e-05,
      "loss": 0.3727,
      "step": 1240
    },
    {
      "epoch": 0.45,
      "learning_rate": 3.900480059084195e-05,
      "loss": 0.2173,
      "step": 1241
    },
    {
      "epoch": 0.45,
      "learning_rate": 3.8995568685376664e-05,
      "loss": 0.2781,
      "step": 1242
    },
    {
      "epoch": 0.45,
      "learning_rate": 3.8986336779911375e-05,
      "loss": 0.0202,
      "step": 1243
    },
    {
      "epoch": 0.46,
      "learning_rate": 3.8977104874446085e-05,
      "loss": 0.1161,
      "step": 1244
    },
    {
      "epoch": 0.46,
      "learning_rate": 3.89678729689808e-05,
      "loss": 0.0191,
      "step": 1245
    },
    {
      "epoch": 0.46,
      "learning_rate": 3.895864106351551e-05,
      "loss": 0.0222,
      "step": 1246
    },
    {
      "epoch": 0.46,
      "learning_rate": 3.8949409158050224e-05,
      "loss": 0.5295,
      "step": 1247
    },
    {
      "epoch": 0.46,
      "learning_rate": 3.8940177252584934e-05,
      "loss": 0.0835,
      "step": 1248
    },
    {
      "epoch": 0.46,
      "learning_rate": 3.8930945347119645e-05,
      "loss": 0.0453,
      "step": 1249
    },
    {
      "epoch": 0.46,
      "learning_rate": 3.892171344165436e-05,
      "loss": 0.4145,
      "step": 1250
    },
    {
      "epoch": 0.46,
      "learning_rate": 3.891248153618907e-05,
      "loss": 0.0298,
      "step": 1251
    },
    {
      "epoch": 0.46,
      "learning_rate": 3.890324963072378e-05,
      "loss": 0.0149,
      "step": 1252
    },
    {
      "epoch": 0.46,
      "learning_rate": 3.8894017725258494e-05,
      "loss": 0.019,
      "step": 1253
    },
    {
      "epoch": 0.46,
      "learning_rate": 3.8884785819793204e-05,
      "loss": 0.2671,
      "step": 1254
    },
    {
      "epoch": 0.46,
      "learning_rate": 3.8875553914327915e-05,
      "loss": 0.2913,
      "step": 1255
    },
    {
      "epoch": 0.46,
      "learning_rate": 3.886632200886263e-05,
      "loss": 0.0237,
      "step": 1256
    },
    {
      "epoch": 0.46,
      "learning_rate": 3.885709010339734e-05,
      "loss": 0.1418,
      "step": 1257
    },
    {
      "epoch": 0.46,
      "learning_rate": 3.884785819793206e-05,
      "loss": 0.0799,
      "step": 1258
    },
    {
      "epoch": 0.46,
      "learning_rate": 3.883862629246677e-05,
      "loss": 0.0097,
      "step": 1259
    },
    {
      "epoch": 0.46,
      "learning_rate": 3.8829394387001474e-05,
      "loss": 0.0327,
      "step": 1260
    },
    {
      "epoch": 0.46,
      "learning_rate": 3.882016248153619e-05,
      "loss": 0.0079,
      "step": 1261
    },
    {
      "epoch": 0.46,
      "learning_rate": 3.88109305760709e-05,
      "loss": 0.0076,
      "step": 1262
    },
    {
      "epoch": 0.46,
      "learning_rate": 3.880169867060561e-05,
      "loss": 0.0261,
      "step": 1263
    },
    {
      "epoch": 0.46,
      "learning_rate": 3.879246676514033e-05,
      "loss": 0.4356,
      "step": 1264
    },
    {
      "epoch": 0.46,
      "learning_rate": 3.878323485967504e-05,
      "loss": 0.1994,
      "step": 1265
    },
    {
      "epoch": 0.46,
      "learning_rate": 3.877400295420975e-05,
      "loss": 0.0153,
      "step": 1266
    },
    {
      "epoch": 0.46,
      "learning_rate": 3.876477104874446e-05,
      "loss": 0.027,
      "step": 1267
    },
    {
      "epoch": 0.46,
      "learning_rate": 3.875553914327917e-05,
      "loss": 0.0113,
      "step": 1268
    },
    {
      "epoch": 0.46,
      "learning_rate": 3.874630723781389e-05,
      "loss": 0.013,
      "step": 1269
    },
    {
      "epoch": 0.46,
      "learning_rate": 3.87370753323486e-05,
      "loss": 0.027,
      "step": 1270
    },
    {
      "epoch": 0.47,
      "learning_rate": 3.872784342688331e-05,
      "loss": 0.0093,
      "step": 1271
    },
    {
      "epoch": 0.47,
      "learning_rate": 3.871861152141802e-05,
      "loss": 0.2166,
      "step": 1272
    },
    {
      "epoch": 0.47,
      "learning_rate": 3.870937961595273e-05,
      "loss": 0.1184,
      "step": 1273
    },
    {
      "epoch": 0.47,
      "learning_rate": 3.870014771048744e-05,
      "loss": 0.5101,
      "step": 1274
    },
    {
      "epoch": 0.47,
      "learning_rate": 3.869091580502216e-05,
      "loss": 0.0926,
      "step": 1275
    },
    {
      "epoch": 0.47,
      "learning_rate": 3.868168389955687e-05,
      "loss": 0.0078,
      "step": 1276
    },
    {
      "epoch": 0.47,
      "learning_rate": 3.867245199409158e-05,
      "loss": 0.2731,
      "step": 1277
    },
    {
      "epoch": 0.47,
      "learning_rate": 3.86632200886263e-05,
      "loss": 0.3515,
      "step": 1278
    },
    {
      "epoch": 0.47,
      "learning_rate": 3.8653988183161e-05,
      "loss": 0.0077,
      "step": 1279
    },
    {
      "epoch": 0.47,
      "learning_rate": 3.864475627769572e-05,
      "loss": 0.0083,
      "step": 1280
    },
    {
      "epoch": 0.47,
      "learning_rate": 3.863552437223043e-05,
      "loss": 0.0821,
      "step": 1281
    },
    {
      "epoch": 0.47,
      "learning_rate": 3.862629246676514e-05,
      "loss": 0.0066,
      "step": 1282
    },
    {
      "epoch": 0.47,
      "learning_rate": 3.861706056129986e-05,
      "loss": 0.2614,
      "step": 1283
    },
    {
      "epoch": 0.47,
      "learning_rate": 3.860782865583457e-05,
      "loss": 0.485,
      "step": 1284
    },
    {
      "epoch": 0.47,
      "learning_rate": 3.859859675036928e-05,
      "loss": 0.4464,
      "step": 1285
    },
    {
      "epoch": 0.47,
      "learning_rate": 3.858936484490399e-05,
      "loss": 0.2081,
      "step": 1286
    },
    {
      "epoch": 0.47,
      "learning_rate": 3.85801329394387e-05,
      "loss": 0.0165,
      "step": 1287
    },
    {
      "epoch": 0.47,
      "learning_rate": 3.857090103397342e-05,
      "loss": 0.0434,
      "step": 1288
    },
    {
      "epoch": 0.47,
      "learning_rate": 3.856166912850813e-05,
      "loss": 0.1392,
      "step": 1289
    },
    {
      "epoch": 0.47,
      "learning_rate": 3.855243722304284e-05,
      "loss": 0.0114,
      "step": 1290
    },
    {
      "epoch": 0.47,
      "learning_rate": 3.854320531757755e-05,
      "loss": 0.0335,
      "step": 1291
    },
    {
      "epoch": 0.47,
      "learning_rate": 3.853397341211226e-05,
      "loss": 0.2187,
      "step": 1292
    },
    {
      "epoch": 0.47,
      "learning_rate": 3.852474150664697e-05,
      "loss": 0.224,
      "step": 1293
    },
    {
      "epoch": 0.47,
      "learning_rate": 3.851550960118169e-05,
      "loss": 0.2955,
      "step": 1294
    },
    {
      "epoch": 0.47,
      "learning_rate": 3.85062776957164e-05,
      "loss": 0.1719,
      "step": 1295
    },
    {
      "epoch": 0.47,
      "learning_rate": 3.849704579025111e-05,
      "loss": 0.0276,
      "step": 1296
    },
    {
      "epoch": 0.47,
      "learning_rate": 3.8487813884785826e-05,
      "loss": 0.257,
      "step": 1297
    },
    {
      "epoch": 0.47,
      "learning_rate": 3.847858197932053e-05,
      "loss": 0.0467,
      "step": 1298
    },
    {
      "epoch": 0.48,
      "learning_rate": 3.846935007385525e-05,
      "loss": 0.0272,
      "step": 1299
    },
    {
      "epoch": 0.48,
      "learning_rate": 3.846011816838996e-05,
      "loss": 0.0103,
      "step": 1300
    },
    {
      "epoch": 0.48,
      "learning_rate": 3.845088626292467e-05,
      "loss": 0.0194,
      "step": 1301
    },
    {
      "epoch": 0.48,
      "learning_rate": 3.8441654357459385e-05,
      "loss": 0.0255,
      "step": 1302
    },
    {
      "epoch": 0.48,
      "learning_rate": 3.8432422451994096e-05,
      "loss": 0.0179,
      "step": 1303
    },
    {
      "epoch": 0.48,
      "learning_rate": 3.8423190546528806e-05,
      "loss": 0.0269,
      "step": 1304
    },
    {
      "epoch": 0.48,
      "learning_rate": 3.841395864106352e-05,
      "loss": 0.4592,
      "step": 1305
    },
    {
      "epoch": 0.48,
      "learning_rate": 3.840472673559823e-05,
      "loss": 0.0045,
      "step": 1306
    },
    {
      "epoch": 0.48,
      "learning_rate": 3.839549483013294e-05,
      "loss": 0.3065,
      "step": 1307
    },
    {
      "epoch": 0.48,
      "learning_rate": 3.8386262924667655e-05,
      "loss": 0.1879,
      "step": 1308
    },
    {
      "epoch": 0.48,
      "learning_rate": 3.8377031019202366e-05,
      "loss": 0.2709,
      "step": 1309
    },
    {
      "epoch": 0.48,
      "learning_rate": 3.8367799113737076e-05,
      "loss": 0.1863,
      "step": 1310
    },
    {
      "epoch": 0.48,
      "learning_rate": 3.835856720827179e-05,
      "loss": 0.2238,
      "step": 1311
    },
    {
      "epoch": 0.48,
      "learning_rate": 3.83493353028065e-05,
      "loss": 0.5957,
      "step": 1312
    },
    {
      "epoch": 0.48,
      "learning_rate": 3.8340103397341215e-05,
      "loss": 0.0994,
      "step": 1313
    },
    {
      "epoch": 0.48,
      "learning_rate": 3.8330871491875925e-05,
      "loss": 0.085,
      "step": 1314
    },
    {
      "epoch": 0.48,
      "learning_rate": 3.8321639586410636e-05,
      "loss": 0.1163,
      "step": 1315
    },
    {
      "epoch": 0.48,
      "learning_rate": 3.831240768094535e-05,
      "loss": 0.1443,
      "step": 1316
    },
    {
      "epoch": 0.48,
      "learning_rate": 3.830317577548006e-05,
      "loss": 0.0058,
      "step": 1317
    },
    {
      "epoch": 0.48,
      "learning_rate": 3.829394387001477e-05,
      "loss": 0.0075,
      "step": 1318
    },
    {
      "epoch": 0.48,
      "learning_rate": 3.8284711964549485e-05,
      "loss": 0.0587,
      "step": 1319
    },
    {
      "epoch": 0.48,
      "learning_rate": 3.8275480059084195e-05,
      "loss": 0.0196,
      "step": 1320
    },
    {
      "epoch": 0.48,
      "learning_rate": 3.826624815361891e-05,
      "loss": 0.0149,
      "step": 1321
    },
    {
      "epoch": 0.48,
      "learning_rate": 3.825701624815362e-05,
      "loss": 0.4112,
      "step": 1322
    },
    {
      "epoch": 0.48,
      "learning_rate": 3.8247784342688334e-05,
      "loss": 0.217,
      "step": 1323
    },
    {
      "epoch": 0.48,
      "learning_rate": 3.8238552437223044e-05,
      "loss": 0.0119,
      "step": 1324
    },
    {
      "epoch": 0.48,
      "learning_rate": 3.8229320531757755e-05,
      "loss": 0.1952,
      "step": 1325
    },
    {
      "epoch": 0.49,
      "learning_rate": 3.8220088626292465e-05,
      "loss": 0.0141,
      "step": 1326
    },
    {
      "epoch": 0.49,
      "learning_rate": 3.821085672082718e-05,
      "loss": 0.0364,
      "step": 1327
    },
    {
      "epoch": 0.49,
      "learning_rate": 3.820162481536189e-05,
      "loss": 0.091,
      "step": 1328
    },
    {
      "epoch": 0.49,
      "learning_rate": 3.8192392909896604e-05,
      "loss": 0.2586,
      "step": 1329
    },
    {
      "epoch": 0.49,
      "learning_rate": 3.8183161004431314e-05,
      "loss": 0.0101,
      "step": 1330
    },
    {
      "epoch": 0.49,
      "learning_rate": 3.8173929098966025e-05,
      "loss": 0.171,
      "step": 1331
    },
    {
      "epoch": 0.49,
      "learning_rate": 3.816469719350074e-05,
      "loss": 0.3094,
      "step": 1332
    },
    {
      "epoch": 0.49,
      "learning_rate": 3.815546528803545e-05,
      "loss": 0.0036,
      "step": 1333
    },
    {
      "epoch": 0.49,
      "learning_rate": 3.8146233382570163e-05,
      "loss": 0.0074,
      "step": 1334
    },
    {
      "epoch": 0.49,
      "learning_rate": 3.813700147710488e-05,
      "loss": 0.3153,
      "step": 1335
    },
    {
      "epoch": 0.49,
      "learning_rate": 3.8127769571639585e-05,
      "loss": 0.0091,
      "step": 1336
    },
    {
      "epoch": 0.49,
      "learning_rate": 3.8118537666174295e-05,
      "loss": 0.008,
      "step": 1337
    },
    {
      "epoch": 0.49,
      "learning_rate": 3.810930576070901e-05,
      "loss": 0.2204,
      "step": 1338
    },
    {
      "epoch": 0.49,
      "learning_rate": 3.810007385524372e-05,
      "loss": 0.0971,
      "step": 1339
    },
    {
      "epoch": 0.49,
      "learning_rate": 3.809084194977844e-05,
      "loss": 0.0098,
      "step": 1340
    },
    {
      "epoch": 0.49,
      "learning_rate": 3.808161004431315e-05,
      "loss": 0.0131,
      "step": 1341
    },
    {
      "epoch": 0.49,
      "learning_rate": 3.807237813884786e-05,
      "loss": 0.2129,
      "step": 1342
    },
    {
      "epoch": 0.49,
      "learning_rate": 3.806314623338257e-05,
      "loss": 0.0167,
      "step": 1343
    },
    {
      "epoch": 0.49,
      "learning_rate": 3.805391432791728e-05,
      "loss": 0.4568,
      "step": 1344
    },
    {
      "epoch": 0.49,
      "learning_rate": 3.804468242245199e-05,
      "loss": 0.0147,
      "step": 1345
    },
    {
      "epoch": 0.49,
      "learning_rate": 3.803545051698671e-05,
      "loss": 0.0099,
      "step": 1346
    },
    {
      "epoch": 0.49,
      "learning_rate": 3.802621861152142e-05,
      "loss": 0.0136,
      "step": 1347
    },
    {
      "epoch": 0.49,
      "learning_rate": 3.801698670605613e-05,
      "loss": 0.0623,
      "step": 1348
    },
    {
      "epoch": 0.49,
      "learning_rate": 3.800775480059084e-05,
      "loss": 0.4716,
      "step": 1349
    },
    {
      "epoch": 0.49,
      "learning_rate": 3.799852289512555e-05,
      "loss": 0.0193,
      "step": 1350
    },
    {
      "epoch": 0.49,
      "learning_rate": 3.798929098966027e-05,
      "loss": 0.0458,
      "step": 1351
    },
    {
      "epoch": 0.49,
      "learning_rate": 3.798005908419498e-05,
      "loss": 0.0116,
      "step": 1352
    },
    {
      "epoch": 0.5,
      "learning_rate": 3.797082717872969e-05,
      "loss": 0.2771,
      "step": 1353
    },
    {
      "epoch": 0.5,
      "learning_rate": 3.796159527326441e-05,
      "loss": 0.0058,
      "step": 1354
    },
    {
      "epoch": 0.5,
      "learning_rate": 3.795236336779911e-05,
      "loss": 0.0675,
      "step": 1355
    },
    {
      "epoch": 0.5,
      "learning_rate": 3.794313146233382e-05,
      "loss": 0.009,
      "step": 1356
    },
    {
      "epoch": 0.5,
      "learning_rate": 3.793389955686854e-05,
      "loss": 0.004,
      "step": 1357
    },
    {
      "epoch": 0.5,
      "learning_rate": 3.792466765140325e-05,
      "loss": 0.5432,
      "step": 1358
    },
    {
      "epoch": 0.5,
      "learning_rate": 3.791543574593797e-05,
      "loss": 0.0081,
      "step": 1359
    },
    {
      "epoch": 0.5,
      "learning_rate": 3.790620384047268e-05,
      "loss": 0.1838,
      "step": 1360
    },
    {
      "epoch": 0.5,
      "learning_rate": 3.789697193500739e-05,
      "loss": 0.199,
      "step": 1361
    },
    {
      "epoch": 0.5,
      "learning_rate": 3.78877400295421e-05,
      "loss": 0.0233,
      "step": 1362
    },
    {
      "epoch": 0.5,
      "learning_rate": 3.787850812407681e-05,
      "loss": 0.0044,
      "step": 1363
    },
    {
      "epoch": 0.5,
      "learning_rate": 3.786927621861152e-05,
      "loss": 0.0503,
      "step": 1364
    },
    {
      "epoch": 0.5,
      "learning_rate": 3.786004431314624e-05,
      "loss": 0.14,
      "step": 1365
    },
    {
      "epoch": 0.5,
      "learning_rate": 3.785081240768095e-05,
      "loss": 0.0523,
      "step": 1366
    },
    {
      "epoch": 0.5,
      "learning_rate": 3.784158050221566e-05,
      "loss": 0.0083,
      "step": 1367
    },
    {
      "epoch": 0.5,
      "learning_rate": 3.783234859675037e-05,
      "loss": 0.093,
      "step": 1368
    },
    {
      "epoch": 0.5,
      "learning_rate": 3.782311669128508e-05,
      "loss": 0.049,
      "step": 1369
    },
    {
      "epoch": 0.5,
      "learning_rate": 3.78138847858198e-05,
      "loss": 0.0107,
      "step": 1370
    },
    {
      "epoch": 0.5,
      "learning_rate": 3.780465288035451e-05,
      "loss": 0.3482,
      "step": 1371
    },
    {
      "epoch": 0.5,
      "learning_rate": 3.779542097488922e-05,
      "loss": 0.0076,
      "step": 1372
    },
    {
      "epoch": 0.5,
      "learning_rate": 3.7786189069423936e-05,
      "loss": 0.299,
      "step": 1373
    },
    {
      "epoch": 0.5,
      "learning_rate": 3.777695716395864e-05,
      "loss": 0.2879,
      "step": 1374
    },
    {
      "epoch": 0.5,
      "learning_rate": 3.776772525849335e-05,
      "loss": 0.004,
      "step": 1375
    },
    {
      "epoch": 0.5,
      "learning_rate": 3.775849335302807e-05,
      "loss": 0.0276,
      "step": 1376
    },
    {
      "epoch": 0.5,
      "learning_rate": 3.774926144756278e-05,
      "loss": 0.0034,
      "step": 1377
    },
    {
      "epoch": 0.5,
      "learning_rate": 3.774002954209749e-05,
      "loss": 0.4207,
      "step": 1378
    },
    {
      "epoch": 0.5,
      "learning_rate": 3.7730797636632206e-05,
      "loss": 0.0056,
      "step": 1379
    },
    {
      "epoch": 0.5,
      "learning_rate": 3.7721565731166916e-05,
      "loss": 0.0066,
      "step": 1380
    },
    {
      "epoch": 0.51,
      "learning_rate": 3.771233382570163e-05,
      "loss": 0.0037,
      "step": 1381
    },
    {
      "epoch": 0.51,
      "learning_rate": 3.770310192023634e-05,
      "loss": 0.0052,
      "step": 1382
    },
    {
      "epoch": 0.51,
      "learning_rate": 3.769387001477105e-05,
      "loss": 0.0438,
      "step": 1383
    },
    {
      "epoch": 0.51,
      "learning_rate": 3.7684638109305765e-05,
      "loss": 0.0036,
      "step": 1384
    },
    {
      "epoch": 0.51,
      "learning_rate": 3.7675406203840476e-05,
      "loss": 0.2312,
      "step": 1385
    },
    {
      "epoch": 0.51,
      "learning_rate": 3.7666174298375187e-05,
      "loss": 0.0153,
      "step": 1386
    },
    {
      "epoch": 0.51,
      "learning_rate": 3.76569423929099e-05,
      "loss": 0.0034,
      "step": 1387
    },
    {
      "epoch": 0.51,
      "learning_rate": 3.764771048744461e-05,
      "loss": 0.0047,
      "step": 1388
    },
    {
      "epoch": 0.51,
      "learning_rate": 3.7638478581979325e-05,
      "loss": 0.0042,
      "step": 1389
    },
    {
      "epoch": 0.51,
      "learning_rate": 3.7629246676514035e-05,
      "loss": 0.0257,
      "step": 1390
    },
    {
      "epoch": 0.51,
      "learning_rate": 3.7620014771048746e-05,
      "loss": 0.0032,
      "step": 1391
    },
    {
      "epoch": 0.51,
      "learning_rate": 3.761078286558346e-05,
      "loss": 0.1491,
      "step": 1392
    },
    {
      "epoch": 0.51,
      "learning_rate": 3.760155096011817e-05,
      "loss": 0.0038,
      "step": 1393
    },
    {
      "epoch": 0.51,
      "learning_rate": 3.759231905465288e-05,
      "loss": 0.306,
      "step": 1394
    },
    {
      "epoch": 0.51,
      "learning_rate": 3.7583087149187595e-05,
      "loss": 0.0069,
      "step": 1395
    },
    {
      "epoch": 0.51,
      "learning_rate": 3.7573855243722306e-05,
      "loss": 0.174,
      "step": 1396
    },
    {
      "epoch": 0.51,
      "learning_rate": 3.7564623338257016e-05,
      "loss": 0.5432,
      "step": 1397
    },
    {
      "epoch": 0.51,
      "learning_rate": 3.7555391432791733e-05,
      "loss": 0.0095,
      "step": 1398
    },
    {
      "epoch": 0.51,
      "learning_rate": 3.7546159527326444e-05,
      "loss": 0.0067,
      "step": 1399
    },
    {
      "epoch": 0.51,
      "learning_rate": 3.7536927621861155e-05,
      "loss": 0.007,
      "step": 1400
    },
    {
      "epoch": 0.51,
      "learning_rate": 3.7527695716395865e-05,
      "loss": 0.0022,
      "step": 1401
    },
    {
      "epoch": 0.51,
      "learning_rate": 3.7518463810930576e-05,
      "loss": 0.0036,
      "step": 1402
    },
    {
      "epoch": 0.51,
      "learning_rate": 3.750923190546529e-05,
      "loss": 0.0042,
      "step": 1403
    },
    {
      "epoch": 0.51,
      "learning_rate": 3.7500000000000003e-05,
      "loss": 0.1525,
      "step": 1404
    },
    {
      "epoch": 0.51,
      "learning_rate": 3.7490768094534714e-05,
      "loss": 0.0049,
      "step": 1405
    },
    {
      "epoch": 0.51,
      "learning_rate": 3.7481536189069425e-05,
      "loss": 0.3763,
      "step": 1406
    },
    {
      "epoch": 0.51,
      "learning_rate": 3.7472304283604135e-05,
      "loss": 0.249,
      "step": 1407
    },
    {
      "epoch": 0.52,
      "learning_rate": 3.7463072378138846e-05,
      "loss": 0.0051,
      "step": 1408
    },
    {
      "epoch": 0.52,
      "learning_rate": 3.745384047267356e-05,
      "loss": 0.0022,
      "step": 1409
    },
    {
      "epoch": 0.52,
      "learning_rate": 3.7444608567208274e-05,
      "loss": 0.0131,
      "step": 1410
    },
    {
      "epoch": 0.52,
      "learning_rate": 3.743537666174299e-05,
      "loss": 0.1274,
      "step": 1411
    },
    {
      "epoch": 0.52,
      "learning_rate": 3.7426144756277695e-05,
      "loss": 0.4141,
      "step": 1412
    },
    {
      "epoch": 0.52,
      "learning_rate": 3.7416912850812405e-05,
      "loss": 0.0022,
      "step": 1413
    },
    {
      "epoch": 0.52,
      "learning_rate": 3.740768094534712e-05,
      "loss": 0.4521,
      "step": 1414
    },
    {
      "epoch": 0.52,
      "learning_rate": 3.739844903988183e-05,
      "loss": 0.4417,
      "step": 1415
    },
    {
      "epoch": 0.52,
      "learning_rate": 3.7389217134416544e-05,
      "loss": 0.0185,
      "step": 1416
    },
    {
      "epoch": 0.52,
      "learning_rate": 3.737998522895126e-05,
      "loss": 0.019,
      "step": 1417
    },
    {
      "epoch": 0.52,
      "learning_rate": 3.737075332348597e-05,
      "loss": 0.0875,
      "step": 1418
    },
    {
      "epoch": 0.52,
      "learning_rate": 3.7361521418020675e-05,
      "loss": 0.6341,
      "step": 1419
    },
    {
      "epoch": 0.52,
      "learning_rate": 3.735228951255539e-05,
      "loss": 0.1545,
      "step": 1420
    },
    {
      "epoch": 0.52,
      "learning_rate": 3.73430576070901e-05,
      "loss": 0.003,
      "step": 1421
    },
    {
      "epoch": 0.52,
      "learning_rate": 3.733382570162482e-05,
      "loss": 0.0259,
      "step": 1422
    },
    {
      "epoch": 0.52,
      "learning_rate": 3.732459379615953e-05,
      "loss": 0.2033,
      "step": 1423
    },
    {
      "epoch": 0.52,
      "learning_rate": 3.731536189069424e-05,
      "loss": 0.0818,
      "step": 1424
    },
    {
      "epoch": 0.52,
      "learning_rate": 3.730612998522895e-05,
      "loss": 0.0029,
      "step": 1425
    },
    {
      "epoch": 0.52,
      "learning_rate": 3.729689807976366e-05,
      "loss": 0.4007,
      "step": 1426
    },
    {
      "epoch": 0.52,
      "learning_rate": 3.728766617429837e-05,
      "loss": 0.3301,
      "step": 1427
    },
    {
      "epoch": 0.52,
      "learning_rate": 3.727843426883309e-05,
      "loss": 0.0023,
      "step": 1428
    },
    {
      "epoch": 0.52,
      "learning_rate": 3.72692023633678e-05,
      "loss": 0.0019,
      "step": 1429
    },
    {
      "epoch": 0.52,
      "learning_rate": 3.725997045790252e-05,
      "loss": 0.3479,
      "step": 1430
    },
    {
      "epoch": 0.52,
      "learning_rate": 3.725073855243722e-05,
      "loss": 0.1241,
      "step": 1431
    },
    {
      "epoch": 0.52,
      "learning_rate": 3.724150664697193e-05,
      "loss": 0.0082,
      "step": 1432
    },
    {
      "epoch": 0.52,
      "learning_rate": 3.723227474150665e-05,
      "loss": 0.4145,
      "step": 1433
    },
    {
      "epoch": 0.52,
      "learning_rate": 3.722304283604136e-05,
      "loss": 0.1708,
      "step": 1434
    },
    {
      "epoch": 0.53,
      "learning_rate": 3.721381093057607e-05,
      "loss": 0.0037,
      "step": 1435
    },
    {
      "epoch": 0.53,
      "learning_rate": 3.720457902511079e-05,
      "loss": 0.018,
      "step": 1436
    },
    {
      "epoch": 0.53,
      "learning_rate": 3.71953471196455e-05,
      "loss": 0.1789,
      "step": 1437
    },
    {
      "epoch": 0.53,
      "learning_rate": 3.71861152141802e-05,
      "loss": 0.0276,
      "step": 1438
    },
    {
      "epoch": 0.53,
      "learning_rate": 3.717688330871492e-05,
      "loss": 0.006,
      "step": 1439
    },
    {
      "epoch": 0.53,
      "learning_rate": 3.716765140324963e-05,
      "loss": 0.0053,
      "step": 1440
    },
    {
      "epoch": 0.53,
      "learning_rate": 3.715841949778435e-05,
      "loss": 0.2272,
      "step": 1441
    },
    {
      "epoch": 0.53,
      "learning_rate": 3.714918759231906e-05,
      "loss": 0.2667,
      "step": 1442
    },
    {
      "epoch": 0.53,
      "learning_rate": 3.713995568685377e-05,
      "loss": 0.2049,
      "step": 1443
    },
    {
      "epoch": 0.53,
      "learning_rate": 3.713072378138848e-05,
      "loss": 0.2252,
      "step": 1444
    },
    {
      "epoch": 0.53,
      "learning_rate": 3.712149187592319e-05,
      "loss": 0.0024,
      "step": 1445
    },
    {
      "epoch": 0.53,
      "learning_rate": 3.71122599704579e-05,
      "loss": 0.0038,
      "step": 1446
    },
    {
      "epoch": 0.53,
      "learning_rate": 3.710302806499262e-05,
      "loss": 0.0048,
      "step": 1447
    },
    {
      "epoch": 0.53,
      "learning_rate": 3.709379615952733e-05,
      "loss": 0.0016,
      "step": 1448
    },
    {
      "epoch": 0.53,
      "learning_rate": 3.708456425406204e-05,
      "loss": 0.061,
      "step": 1449
    },
    {
      "epoch": 0.53,
      "learning_rate": 3.707533234859675e-05,
      "loss": 0.0202,
      "step": 1450
    },
    {
      "epoch": 0.53,
      "learning_rate": 3.706610044313146e-05,
      "loss": 0.027,
      "step": 1451
    },
    {
      "epoch": 0.53,
      "learning_rate": 3.705686853766618e-05,
      "loss": 0.0038,
      "step": 1452
    },
    {
      "epoch": 0.53,
      "learning_rate": 3.704763663220089e-05,
      "loss": 0.0975,
      "step": 1453
    },
    {
      "epoch": 0.53,
      "learning_rate": 3.70384047267356e-05,
      "loss": 0.011,
      "step": 1454
    },
    {
      "epoch": 0.53,
      "learning_rate": 3.7029172821270316e-05,
      "loss": 0.1517,
      "step": 1455
    },
    {
      "epoch": 0.53,
      "learning_rate": 3.7019940915805027e-05,
      "loss": 0.3696,
      "step": 1456
    },
    {
      "epoch": 0.53,
      "learning_rate": 3.701070901033973e-05,
      "loss": 0.0203,
      "step": 1457
    },
    {
      "epoch": 0.53,
      "learning_rate": 3.700147710487445e-05,
      "loss": 0.1267,
      "step": 1458
    },
    {
      "epoch": 0.53,
      "learning_rate": 3.699224519940916e-05,
      "loss": 0.0037,
      "step": 1459
    },
    {
      "epoch": 0.53,
      "learning_rate": 3.6983013293943876e-05,
      "loss": 0.002,
      "step": 1460
    },
    {
      "epoch": 0.53,
      "learning_rate": 3.6973781388478586e-05,
      "loss": 0.0025,
      "step": 1461
    },
    {
      "epoch": 0.53,
      "learning_rate": 3.69645494830133e-05,
      "loss": 0.0047,
      "step": 1462
    },
    {
      "epoch": 0.54,
      "learning_rate": 3.695531757754801e-05,
      "loss": 0.0023,
      "step": 1463
    },
    {
      "epoch": 0.54,
      "learning_rate": 3.694608567208272e-05,
      "loss": 0.0079,
      "step": 1464
    },
    {
      "epoch": 0.54,
      "learning_rate": 3.693685376661743e-05,
      "loss": 0.4306,
      "step": 1465
    },
    {
      "epoch": 0.54,
      "learning_rate": 3.6927621861152146e-05,
      "loss": 0.4872,
      "step": 1466
    },
    {
      "epoch": 0.54,
      "learning_rate": 3.6918389955686856e-05,
      "loss": 0.0075,
      "step": 1467
    },
    {
      "epoch": 0.54,
      "learning_rate": 3.690915805022157e-05,
      "loss": 0.1472,
      "step": 1468
    },
    {
      "epoch": 0.54,
      "learning_rate": 3.689992614475628e-05,
      "loss": 0.3181,
      "step": 1469
    },
    {
      "epoch": 0.54,
      "learning_rate": 3.689069423929099e-05,
      "loss": 0.0228,
      "step": 1470
    },
    {
      "epoch": 0.54,
      "learning_rate": 3.6881462333825705e-05,
      "loss": 0.0428,
      "step": 1471
    },
    {
      "epoch": 0.54,
      "learning_rate": 3.6872230428360416e-05,
      "loss": 0.0038,
      "step": 1472
    },
    {
      "epoch": 0.54,
      "learning_rate": 3.6862998522895126e-05,
      "loss": 0.0091,
      "step": 1473
    },
    {
      "epoch": 0.54,
      "learning_rate": 3.6853766617429844e-05,
      "loss": 0.0209,
      "step": 1474
    },
    {
      "epoch": 0.54,
      "learning_rate": 3.6844534711964554e-05,
      "loss": 0.1103,
      "step": 1475
    },
    {
      "epoch": 0.54,
      "learning_rate": 3.683530280649926e-05,
      "loss": 0.0075,
      "step": 1476
    },
    {
      "epoch": 0.54,
      "learning_rate": 3.6826070901033975e-05,
      "loss": 0.0158,
      "step": 1477
    },
    {
      "epoch": 0.54,
      "learning_rate": 3.6816838995568686e-05,
      "loss": 0.122,
      "step": 1478
    },
    {
      "epoch": 0.54,
      "learning_rate": 3.6807607090103396e-05,
      "loss": 0.0048,
      "step": 1479
    },
    {
      "epoch": 0.54,
      "learning_rate": 3.6798375184638114e-05,
      "loss": 0.3,
      "step": 1480
    },
    {
      "epoch": 0.54,
      "learning_rate": 3.6789143279172824e-05,
      "loss": 0.0051,
      "step": 1481
    },
    {
      "epoch": 0.54,
      "learning_rate": 3.6779911373707535e-05,
      "loss": 0.003,
      "step": 1482
    },
    {
      "epoch": 0.54,
      "learning_rate": 3.6770679468242245e-05,
      "loss": 0.0035,
      "step": 1483
    },
    {
      "epoch": 0.54,
      "learning_rate": 3.6761447562776956e-05,
      "loss": 0.4367,
      "step": 1484
    },
    {
      "epoch": 0.54,
      "learning_rate": 3.675221565731167e-05,
      "loss": 0.0051,
      "step": 1485
    },
    {
      "epoch": 0.54,
      "learning_rate": 3.6742983751846384e-05,
      "loss": 0.1593,
      "step": 1486
    },
    {
      "epoch": 0.54,
      "learning_rate": 3.6733751846381094e-05,
      "loss": 0.1618,
      "step": 1487
    },
    {
      "epoch": 0.54,
      "learning_rate": 3.6724519940915805e-05,
      "loss": 0.3911,
      "step": 1488
    },
    {
      "epoch": 0.54,
      "learning_rate": 3.6715288035450515e-05,
      "loss": 0.2104,
      "step": 1489
    },
    {
      "epoch": 0.55,
      "learning_rate": 3.670605612998523e-05,
      "loss": 0.0123,
      "step": 1490
    },
    {
      "epoch": 0.55,
      "learning_rate": 3.669682422451994e-05,
      "loss": 0.0642,
      "step": 1491
    },
    {
      "epoch": 0.55,
      "learning_rate": 3.6687592319054654e-05,
      "loss": 0.3308,
      "step": 1492
    },
    {
      "epoch": 0.55,
      "learning_rate": 3.667836041358937e-05,
      "loss": 0.2146,
      "step": 1493
    },
    {
      "epoch": 0.55,
      "learning_rate": 3.666912850812408e-05,
      "loss": 0.2373,
      "step": 1494
    },
    {
      "epoch": 0.55,
      "learning_rate": 3.6659896602658785e-05,
      "loss": 0.0146,
      "step": 1495
    },
    {
      "epoch": 0.55,
      "learning_rate": 3.66506646971935e-05,
      "loss": 0.4017,
      "step": 1496
    },
    {
      "epoch": 0.55,
      "learning_rate": 3.664143279172821e-05,
      "loss": 0.0803,
      "step": 1497
    },
    {
      "epoch": 0.55,
      "learning_rate": 3.6632200886262924e-05,
      "loss": 0.1962,
      "step": 1498
    },
    {
      "epoch": 0.55,
      "learning_rate": 3.662296898079764e-05,
      "loss": 0.0453,
      "step": 1499
    },
    {
      "epoch": 0.55,
      "learning_rate": 3.661373707533235e-05,
      "loss": 0.3444,
      "step": 1500
    },
    {
      "epoch": 0.55,
      "learning_rate": 3.660450516986706e-05,
      "loss": 0.0261,
      "step": 1501
    },
    {
      "epoch": 0.55,
      "learning_rate": 3.659527326440177e-05,
      "loss": 0.0438,
      "step": 1502
    },
    {
      "epoch": 0.55,
      "learning_rate": 3.658604135893648e-05,
      "loss": 0.2392,
      "step": 1503
    },
    {
      "epoch": 0.55,
      "learning_rate": 3.65768094534712e-05,
      "loss": 0.0203,
      "step": 1504
    },
    {
      "epoch": 0.55,
      "learning_rate": 3.656757754800591e-05,
      "loss": 0.0504,
      "step": 1505
    },
    {
      "epoch": 0.55,
      "learning_rate": 3.655834564254062e-05,
      "loss": 0.1788,
      "step": 1506
    },
    {
      "epoch": 0.55,
      "learning_rate": 3.654911373707533e-05,
      "loss": 0.0208,
      "step": 1507
    },
    {
      "epoch": 0.55,
      "learning_rate": 3.653988183161004e-05,
      "loss": 0.005,
      "step": 1508
    },
    {
      "epoch": 0.55,
      "learning_rate": 3.6530649926144753e-05,
      "loss": 0.0991,
      "step": 1509
    },
    {
      "epoch": 0.55,
      "learning_rate": 3.652141802067947e-05,
      "loss": 0.1446,
      "step": 1510
    },
    {
      "epoch": 0.55,
      "learning_rate": 3.651218611521418e-05,
      "loss": 0.0219,
      "step": 1511
    },
    {
      "epoch": 0.55,
      "learning_rate": 3.65029542097489e-05,
      "loss": 0.0694,
      "step": 1512
    },
    {
      "epoch": 0.55,
      "learning_rate": 3.649372230428361e-05,
      "loss": 0.1725,
      "step": 1513
    },
    {
      "epoch": 0.55,
      "learning_rate": 3.648449039881831e-05,
      "loss": 0.2085,
      "step": 1514
    },
    {
      "epoch": 0.55,
      "learning_rate": 3.647525849335303e-05,
      "loss": 0.004,
      "step": 1515
    },
    {
      "epoch": 0.55,
      "learning_rate": 3.646602658788774e-05,
      "loss": 0.0116,
      "step": 1516
    },
    {
      "epoch": 0.56,
      "learning_rate": 3.645679468242245e-05,
      "loss": 0.1473,
      "step": 1517
    },
    {
      "epoch": 0.56,
      "learning_rate": 3.644756277695717e-05,
      "loss": 0.1895,
      "step": 1518
    },
    {
      "epoch": 0.56,
      "learning_rate": 3.643833087149188e-05,
      "loss": 0.0114,
      "step": 1519
    },
    {
      "epoch": 0.56,
      "learning_rate": 3.642909896602659e-05,
      "loss": 0.0365,
      "step": 1520
    },
    {
      "epoch": 0.56,
      "learning_rate": 3.64198670605613e-05,
      "loss": 0.0255,
      "step": 1521
    },
    {
      "epoch": 0.56,
      "learning_rate": 3.641063515509601e-05,
      "loss": 0.2643,
      "step": 1522
    },
    {
      "epoch": 0.56,
      "learning_rate": 3.640140324963073e-05,
      "loss": 0.0101,
      "step": 1523
    },
    {
      "epoch": 0.56,
      "learning_rate": 3.639217134416544e-05,
      "loss": 0.0101,
      "step": 1524
    },
    {
      "epoch": 0.56,
      "learning_rate": 3.638293943870015e-05,
      "loss": 0.0077,
      "step": 1525
    },
    {
      "epoch": 0.56,
      "learning_rate": 3.637370753323486e-05,
      "loss": 0.0029,
      "step": 1526
    },
    {
      "epoch": 0.56,
      "learning_rate": 3.636447562776957e-05,
      "loss": 0.0077,
      "step": 1527
    },
    {
      "epoch": 0.56,
      "learning_rate": 3.635524372230428e-05,
      "loss": 0.2245,
      "step": 1528
    },
    {
      "epoch": 0.56,
      "learning_rate": 3.6346011816839e-05,
      "loss": 0.0393,
      "step": 1529
    },
    {
      "epoch": 0.56,
      "learning_rate": 3.633677991137371e-05,
      "loss": 0.4033,
      "step": 1530
    },
    {
      "epoch": 0.56,
      "learning_rate": 3.6327548005908426e-05,
      "loss": 0.0086,
      "step": 1531
    },
    {
      "epoch": 0.56,
      "learning_rate": 3.631831610044314e-05,
      "loss": 0.0045,
      "step": 1532
    },
    {
      "epoch": 0.56,
      "learning_rate": 3.630908419497784e-05,
      "loss": 0.014,
      "step": 1533
    },
    {
      "epoch": 0.56,
      "learning_rate": 3.629985228951256e-05,
      "loss": 0.0027,
      "step": 1534
    },
    {
      "epoch": 0.56,
      "learning_rate": 3.629062038404727e-05,
      "loss": 0.3529,
      "step": 1535
    },
    {
      "epoch": 0.56,
      "learning_rate": 3.628138847858198e-05,
      "loss": 0.1788,
      "step": 1536
    },
    {
      "epoch": 0.56,
      "learning_rate": 3.6272156573116696e-05,
      "loss": 0.0501,
      "step": 1537
    },
    {
      "epoch": 0.56,
      "learning_rate": 3.626292466765141e-05,
      "loss": 0.184,
      "step": 1538
    },
    {
      "epoch": 0.56,
      "learning_rate": 3.625369276218612e-05,
      "loss": 0.4035,
      "step": 1539
    },
    {
      "epoch": 0.56,
      "learning_rate": 3.624446085672083e-05,
      "loss": 0.0041,
      "step": 1540
    },
    {
      "epoch": 0.56,
      "learning_rate": 3.623522895125554e-05,
      "loss": 0.0525,
      "step": 1541
    },
    {
      "epoch": 0.56,
      "learning_rate": 3.6225997045790256e-05,
      "loss": 0.035,
      "step": 1542
    },
    {
      "epoch": 0.56,
      "learning_rate": 3.6216765140324966e-05,
      "loss": 0.3515,
      "step": 1543
    },
    {
      "epoch": 0.56,
      "learning_rate": 3.620753323485968e-05,
      "loss": 0.2732,
      "step": 1544
    },
    {
      "epoch": 0.57,
      "learning_rate": 3.619830132939439e-05,
      "loss": 0.005,
      "step": 1545
    },
    {
      "epoch": 0.57,
      "learning_rate": 3.61890694239291e-05,
      "loss": 0.0151,
      "step": 1546
    },
    {
      "epoch": 0.57,
      "learning_rate": 3.617983751846381e-05,
      "loss": 0.4176,
      "step": 1547
    },
    {
      "epoch": 0.57,
      "learning_rate": 3.6170605612998526e-05,
      "loss": 0.0188,
      "step": 1548
    },
    {
      "epoch": 0.57,
      "learning_rate": 3.6161373707533236e-05,
      "loss": 0.2074,
      "step": 1549
    },
    {
      "epoch": 0.57,
      "learning_rate": 3.615214180206795e-05,
      "loss": 0.0933,
      "step": 1550
    },
    {
      "epoch": 0.57,
      "learning_rate": 3.6142909896602664e-05,
      "loss": 0.2267,
      "step": 1551
    },
    {
      "epoch": 0.57,
      "learning_rate": 3.613367799113737e-05,
      "loss": 0.0084,
      "step": 1552
    },
    {
      "epoch": 0.57,
      "learning_rate": 3.6124446085672085e-05,
      "loss": 0.1287,
      "step": 1553
    },
    {
      "epoch": 0.57,
      "learning_rate": 3.6115214180206796e-05,
      "loss": 0.1219,
      "step": 1554
    },
    {
      "epoch": 0.57,
      "learning_rate": 3.6105982274741506e-05,
      "loss": 0.1175,
      "step": 1555
    },
    {
      "epoch": 0.57,
      "learning_rate": 3.6096750369276224e-05,
      "loss": 0.0063,
      "step": 1556
    },
    {
      "epoch": 0.57,
      "learning_rate": 3.6087518463810934e-05,
      "loss": 0.0037,
      "step": 1557
    },
    {
      "epoch": 0.57,
      "learning_rate": 3.6078286558345645e-05,
      "loss": 0.2515,
      "step": 1558
    },
    {
      "epoch": 0.57,
      "learning_rate": 3.6069054652880355e-05,
      "loss": 0.2724,
      "step": 1559
    },
    {
      "epoch": 0.57,
      "learning_rate": 3.6059822747415066e-05,
      "loss": 0.0577,
      "step": 1560
    },
    {
      "epoch": 0.57,
      "learning_rate": 3.605059084194978e-05,
      "loss": 0.0263,
      "step": 1561
    },
    {
      "epoch": 0.57,
      "learning_rate": 3.6041358936484494e-05,
      "loss": 0.024,
      "step": 1562
    },
    {
      "epoch": 0.57,
      "learning_rate": 3.6032127031019204e-05,
      "loss": 0.0132,
      "step": 1563
    },
    {
      "epoch": 0.57,
      "learning_rate": 3.6022895125553915e-05,
      "loss": 0.2734,
      "step": 1564
    },
    {
      "epoch": 0.57,
      "learning_rate": 3.6013663220088626e-05,
      "loss": 0.0185,
      "step": 1565
    },
    {
      "epoch": 0.57,
      "learning_rate": 3.6004431314623336e-05,
      "loss": 0.6792,
      "step": 1566
    },
    {
      "epoch": 0.57,
      "learning_rate": 3.599519940915805e-05,
      "loss": 0.0052,
      "step": 1567
    },
    {
      "epoch": 0.57,
      "learning_rate": 3.5985967503692764e-05,
      "loss": 0.0082,
      "step": 1568
    },
    {
      "epoch": 0.57,
      "learning_rate": 3.5976735598227474e-05,
      "loss": 0.1501,
      "step": 1569
    },
    {
      "epoch": 0.57,
      "learning_rate": 3.596750369276219e-05,
      "loss": 0.0067,
      "step": 1570
    },
    {
      "epoch": 0.57,
      "learning_rate": 3.5958271787296896e-05,
      "loss": 0.2605,
      "step": 1571
    },
    {
      "epoch": 0.58,
      "learning_rate": 3.594903988183161e-05,
      "loss": 0.0035,
      "step": 1572
    },
    {
      "epoch": 0.58,
      "learning_rate": 3.5939807976366323e-05,
      "loss": 0.0183,
      "step": 1573
    },
    {
      "epoch": 0.58,
      "learning_rate": 3.5930576070901034e-05,
      "loss": 0.5915,
      "step": 1574
    },
    {
      "epoch": 0.58,
      "learning_rate": 3.592134416543575e-05,
      "loss": 0.1669,
      "step": 1575
    },
    {
      "epoch": 0.58,
      "learning_rate": 3.591211225997046e-05,
      "loss": 0.0035,
      "step": 1576
    },
    {
      "epoch": 0.58,
      "learning_rate": 3.590288035450517e-05,
      "loss": 0.0265,
      "step": 1577
    },
    {
      "epoch": 0.58,
      "learning_rate": 3.589364844903988e-05,
      "loss": 0.0162,
      "step": 1578
    },
    {
      "epoch": 0.58,
      "learning_rate": 3.5884416543574594e-05,
      "loss": 0.0509,
      "step": 1579
    },
    {
      "epoch": 0.58,
      "learning_rate": 3.5875184638109304e-05,
      "loss": 0.008,
      "step": 1580
    },
    {
      "epoch": 0.58,
      "learning_rate": 3.586595273264402e-05,
      "loss": 0.0054,
      "step": 1581
    },
    {
      "epoch": 0.58,
      "learning_rate": 3.585672082717873e-05,
      "loss": 0.2958,
      "step": 1582
    },
    {
      "epoch": 0.58,
      "learning_rate": 3.584748892171344e-05,
      "loss": 0.222,
      "step": 1583
    },
    {
      "epoch": 0.58,
      "learning_rate": 3.583825701624815e-05,
      "loss": 0.0041,
      "step": 1584
    },
    {
      "epoch": 0.58,
      "learning_rate": 3.5829025110782864e-05,
      "loss": 0.0035,
      "step": 1585
    },
    {
      "epoch": 0.58,
      "learning_rate": 3.581979320531758e-05,
      "loss": 0.0062,
      "step": 1586
    },
    {
      "epoch": 0.58,
      "learning_rate": 3.581056129985229e-05,
      "loss": 0.2804,
      "step": 1587
    },
    {
      "epoch": 0.58,
      "learning_rate": 3.5801329394387e-05,
      "loss": 0.2268,
      "step": 1588
    },
    {
      "epoch": 0.58,
      "learning_rate": 3.579209748892172e-05,
      "loss": 0.0037,
      "step": 1589
    },
    {
      "epoch": 0.58,
      "learning_rate": 3.578286558345642e-05,
      "loss": 0.0058,
      "step": 1590
    },
    {
      "epoch": 0.58,
      "learning_rate": 3.577363367799114e-05,
      "loss": 0.006,
      "step": 1591
    },
    {
      "epoch": 0.58,
      "learning_rate": 3.576440177252585e-05,
      "loss": 0.3415,
      "step": 1592
    },
    {
      "epoch": 0.58,
      "learning_rate": 3.575516986706056e-05,
      "loss": 0.1931,
      "step": 1593
    },
    {
      "epoch": 0.58,
      "learning_rate": 3.574593796159528e-05,
      "loss": 0.0073,
      "step": 1594
    },
    {
      "epoch": 0.58,
      "learning_rate": 3.573670605612999e-05,
      "loss": 0.0042,
      "step": 1595
    },
    {
      "epoch": 0.58,
      "learning_rate": 3.57274741506647e-05,
      "loss": 0.245,
      "step": 1596
    },
    {
      "epoch": 0.58,
      "learning_rate": 3.571824224519941e-05,
      "loss": 0.0071,
      "step": 1597
    },
    {
      "epoch": 0.58,
      "learning_rate": 3.570901033973412e-05,
      "loss": 0.0415,
      "step": 1598
    },
    {
      "epoch": 0.59,
      "learning_rate": 3.569977843426883e-05,
      "loss": 0.0873,
      "step": 1599
    },
    {
      "epoch": 0.59,
      "learning_rate": 3.569054652880355e-05,
      "loss": 0.0034,
      "step": 1600
    },
    {
      "epoch": 0.59,
      "learning_rate": 3.568131462333826e-05,
      "loss": 0.0391,
      "step": 1601
    },
    {
      "epoch": 0.59,
      "learning_rate": 3.567208271787297e-05,
      "loss": 0.0027,
      "step": 1602
    },
    {
      "epoch": 0.59,
      "learning_rate": 3.566285081240768e-05,
      "loss": 0.1675,
      "step": 1603
    },
    {
      "epoch": 0.59,
      "learning_rate": 3.565361890694239e-05,
      "loss": 0.3308,
      "step": 1604
    },
    {
      "epoch": 0.59,
      "learning_rate": 3.564438700147711e-05,
      "loss": 0.0048,
      "step": 1605
    },
    {
      "epoch": 0.59,
      "learning_rate": 3.563515509601182e-05,
      "loss": 0.0066,
      "step": 1606
    },
    {
      "epoch": 0.59,
      "learning_rate": 3.562592319054653e-05,
      "loss": 0.2198,
      "step": 1607
    },
    {
      "epoch": 0.59,
      "learning_rate": 3.561669128508125e-05,
      "loss": 0.2654,
      "step": 1608
    },
    {
      "epoch": 0.59,
      "learning_rate": 3.560745937961595e-05,
      "loss": 0.3402,
      "step": 1609
    },
    {
      "epoch": 0.59,
      "learning_rate": 3.559822747415066e-05,
      "loss": 0.1078,
      "step": 1610
    },
    {
      "epoch": 0.59,
      "learning_rate": 3.558899556868538e-05,
      "loss": 0.0095,
      "step": 1611
    },
    {
      "epoch": 0.59,
      "learning_rate": 3.557976366322009e-05,
      "loss": 0.0079,
      "step": 1612
    },
    {
      "epoch": 0.59,
      "learning_rate": 3.5570531757754806e-05,
      "loss": 0.0094,
      "step": 1613
    },
    {
      "epoch": 0.59,
      "learning_rate": 3.556129985228952e-05,
      "loss": 0.0118,
      "step": 1614
    },
    {
      "epoch": 0.59,
      "learning_rate": 3.555206794682423e-05,
      "loss": 0.0298,
      "step": 1615
    },
    {
      "epoch": 0.59,
      "learning_rate": 3.554283604135894e-05,
      "loss": 0.2324,
      "step": 1616
    },
    {
      "epoch": 0.59,
      "learning_rate": 3.553360413589365e-05,
      "loss": 0.0121,
      "step": 1617
    },
    {
      "epoch": 0.59,
      "learning_rate": 3.552437223042836e-05,
      "loss": 0.0134,
      "step": 1618
    },
    {
      "epoch": 0.59,
      "learning_rate": 3.5515140324963076e-05,
      "loss": 0.0063,
      "step": 1619
    },
    {
      "epoch": 0.59,
      "learning_rate": 3.550590841949779e-05,
      "loss": 0.0226,
      "step": 1620
    },
    {
      "epoch": 0.59,
      "learning_rate": 3.54966765140325e-05,
      "loss": 0.0061,
      "step": 1621
    },
    {
      "epoch": 0.59,
      "learning_rate": 3.548744460856721e-05,
      "loss": 0.0064,
      "step": 1622
    },
    {
      "epoch": 0.59,
      "learning_rate": 3.547821270310192e-05,
      "loss": 0.0174,
      "step": 1623
    },
    {
      "epoch": 0.59,
      "learning_rate": 3.5468980797636636e-05,
      "loss": 0.0088,
      "step": 1624
    },
    {
      "epoch": 0.59,
      "learning_rate": 3.5459748892171347e-05,
      "loss": 0.0153,
      "step": 1625
    },
    {
      "epoch": 0.59,
      "learning_rate": 3.545051698670606e-05,
      "loss": 0.0317,
      "step": 1626
    },
    {
      "epoch": 0.6,
      "learning_rate": 3.5441285081240774e-05,
      "loss": 0.0165,
      "step": 1627
    },
    {
      "epoch": 0.6,
      "learning_rate": 3.543205317577548e-05,
      "loss": 0.0044,
      "step": 1628
    },
    {
      "epoch": 0.6,
      "learning_rate": 3.542282127031019e-05,
      "loss": 0.0599,
      "step": 1629
    },
    {
      "epoch": 0.6,
      "learning_rate": 3.5413589364844906e-05,
      "loss": 0.08,
      "step": 1630
    },
    {
      "epoch": 0.6,
      "learning_rate": 3.5404357459379617e-05,
      "loss": 0.0045,
      "step": 1631
    },
    {
      "epoch": 0.6,
      "learning_rate": 3.5395125553914334e-05,
      "loss": 0.0096,
      "step": 1632
    },
    {
      "epoch": 0.6,
      "learning_rate": 3.5385893648449044e-05,
      "loss": 0.3581,
      "step": 1633
    },
    {
      "epoch": 0.6,
      "learning_rate": 3.5376661742983755e-05,
      "loss": 0.0084,
      "step": 1634
    },
    {
      "epoch": 0.6,
      "learning_rate": 3.5367429837518466e-05,
      "loss": 0.8687,
      "step": 1635
    },
    {
      "epoch": 0.6,
      "learning_rate": 3.5358197932053176e-05,
      "loss": 0.0027,
      "step": 1636
    },
    {
      "epoch": 0.6,
      "learning_rate": 3.534896602658789e-05,
      "loss": 0.0036,
      "step": 1637
    },
    {
      "epoch": 0.6,
      "learning_rate": 3.5339734121122604e-05,
      "loss": 0.2023,
      "step": 1638
    },
    {
      "epoch": 0.6,
      "learning_rate": 3.5330502215657315e-05,
      "loss": 0.0048,
      "step": 1639
    },
    {
      "epoch": 0.6,
      "learning_rate": 3.5321270310192025e-05,
      "loss": 0.0037,
      "step": 1640
    },
    {
      "epoch": 0.6,
      "learning_rate": 3.5312038404726736e-05,
      "loss": 0.9427,
      "step": 1641
    },
    {
      "epoch": 0.6,
      "learning_rate": 3.5302806499261446e-05,
      "loss": 0.0048,
      "step": 1642
    },
    {
      "epoch": 0.6,
      "learning_rate": 3.5293574593796164e-05,
      "loss": 0.678,
      "step": 1643
    },
    {
      "epoch": 0.6,
      "learning_rate": 3.5284342688330874e-05,
      "loss": 0.3142,
      "step": 1644
    },
    {
      "epoch": 0.6,
      "learning_rate": 3.5275110782865585e-05,
      "loss": 0.0453,
      "step": 1645
    },
    {
      "epoch": 0.6,
      "learning_rate": 3.52658788774003e-05,
      "loss": 0.3034,
      "step": 1646
    },
    {
      "epoch": 0.6,
      "learning_rate": 3.5256646971935006e-05,
      "loss": 0.0125,
      "step": 1647
    },
    {
      "epoch": 0.6,
      "learning_rate": 3.5247415066469716e-05,
      "loss": 0.0029,
      "step": 1648
    },
    {
      "epoch": 0.6,
      "learning_rate": 3.5238183161004434e-05,
      "loss": 0.0037,
      "step": 1649
    },
    {
      "epoch": 0.6,
      "learning_rate": 3.5228951255539144e-05,
      "loss": 0.3389,
      "step": 1650
    },
    {
      "epoch": 0.6,
      "learning_rate": 3.5219719350073855e-05,
      "loss": 0.002,
      "step": 1651
    },
    {
      "epoch": 0.6,
      "learning_rate": 3.521048744460857e-05,
      "loss": 0.0062,
      "step": 1652
    },
    {
      "epoch": 0.6,
      "learning_rate": 3.520125553914328e-05,
      "loss": 0.0192,
      "step": 1653
    },
    {
      "epoch": 0.61,
      "learning_rate": 3.519202363367799e-05,
      "loss": 0.291,
      "step": 1654
    },
    {
      "epoch": 0.61,
      "learning_rate": 3.5182791728212704e-05,
      "loss": 0.0026,
      "step": 1655
    },
    {
      "epoch": 0.61,
      "learning_rate": 3.5173559822747414e-05,
      "loss": 0.002,
      "step": 1656
    },
    {
      "epoch": 0.61,
      "learning_rate": 3.516432791728213e-05,
      "loss": 0.033,
      "step": 1657
    },
    {
      "epoch": 0.61,
      "learning_rate": 3.515509601181684e-05,
      "loss": 0.0511,
      "step": 1658
    },
    {
      "epoch": 0.61,
      "learning_rate": 3.514586410635155e-05,
      "loss": 0.0033,
      "step": 1659
    },
    {
      "epoch": 0.61,
      "learning_rate": 3.513663220088626e-05,
      "loss": 0.0149,
      "step": 1660
    },
    {
      "epoch": 0.61,
      "learning_rate": 3.5127400295420974e-05,
      "loss": 0.0313,
      "step": 1661
    },
    {
      "epoch": 0.61,
      "learning_rate": 3.511816838995569e-05,
      "loss": 0.0041,
      "step": 1662
    },
    {
      "epoch": 0.61,
      "learning_rate": 3.51089364844904e-05,
      "loss": 0.0053,
      "step": 1663
    },
    {
      "epoch": 0.61,
      "learning_rate": 3.509970457902511e-05,
      "loss": 0.0037,
      "step": 1664
    },
    {
      "epoch": 0.61,
      "learning_rate": 3.509047267355983e-05,
      "loss": 0.3216,
      "step": 1665
    },
    {
      "epoch": 0.61,
      "learning_rate": 3.508124076809453e-05,
      "loss": 0.0037,
      "step": 1666
    },
    {
      "epoch": 0.61,
      "learning_rate": 3.5072008862629244e-05,
      "loss": 0.0055,
      "step": 1667
    },
    {
      "epoch": 0.61,
      "learning_rate": 3.506277695716396e-05,
      "loss": 0.0673,
      "step": 1668
    },
    {
      "epoch": 0.61,
      "learning_rate": 3.505354505169867e-05,
      "loss": 0.5637,
      "step": 1669
    },
    {
      "epoch": 0.61,
      "learning_rate": 3.504431314623338e-05,
      "loss": 0.2071,
      "step": 1670
    },
    {
      "epoch": 0.61,
      "learning_rate": 3.50350812407681e-05,
      "loss": 0.2626,
      "step": 1671
    },
    {
      "epoch": 0.61,
      "learning_rate": 3.502584933530281e-05,
      "loss": 0.0221,
      "step": 1672
    },
    {
      "epoch": 0.61,
      "learning_rate": 3.501661742983752e-05,
      "loss": 0.3233,
      "step": 1673
    },
    {
      "epoch": 0.61,
      "learning_rate": 3.500738552437223e-05,
      "loss": 0.2114,
      "step": 1674
    },
    {
      "epoch": 0.61,
      "learning_rate": 3.499815361890694e-05,
      "loss": 0.212,
      "step": 1675
    },
    {
      "epoch": 0.61,
      "learning_rate": 3.498892171344166e-05,
      "loss": 0.0039,
      "step": 1676
    },
    {
      "epoch": 0.61,
      "learning_rate": 3.497968980797637e-05,
      "loss": 0.0083,
      "step": 1677
    },
    {
      "epoch": 0.61,
      "learning_rate": 3.497045790251108e-05,
      "loss": 0.0154,
      "step": 1678
    },
    {
      "epoch": 0.61,
      "learning_rate": 3.496122599704579e-05,
      "loss": 0.3823,
      "step": 1679
    },
    {
      "epoch": 0.61,
      "learning_rate": 3.49519940915805e-05,
      "loss": 0.0014,
      "step": 1680
    },
    {
      "epoch": 0.62,
      "learning_rate": 3.494276218611521e-05,
      "loss": 0.1897,
      "step": 1681
    },
    {
      "epoch": 0.62,
      "learning_rate": 3.493353028064993e-05,
      "loss": 0.0041,
      "step": 1682
    },
    {
      "epoch": 0.62,
      "learning_rate": 3.492429837518464e-05,
      "loss": 0.4933,
      "step": 1683
    },
    {
      "epoch": 0.62,
      "learning_rate": 3.491506646971936e-05,
      "loss": 0.0083,
      "step": 1684
    },
    {
      "epoch": 0.62,
      "learning_rate": 3.490583456425406e-05,
      "loss": 0.414,
      "step": 1685
    },
    {
      "epoch": 0.62,
      "learning_rate": 3.489660265878877e-05,
      "loss": 0.0583,
      "step": 1686
    },
    {
      "epoch": 0.62,
      "learning_rate": 3.488737075332349e-05,
      "loss": 0.2824,
      "step": 1687
    },
    {
      "epoch": 0.62,
      "learning_rate": 3.48781388478582e-05,
      "loss": 0.2514,
      "step": 1688
    },
    {
      "epoch": 0.62,
      "learning_rate": 3.486890694239291e-05,
      "loss": 0.222,
      "step": 1689
    },
    {
      "epoch": 0.62,
      "learning_rate": 3.485967503692763e-05,
      "loss": 0.0072,
      "step": 1690
    },
    {
      "epoch": 0.62,
      "learning_rate": 3.485044313146234e-05,
      "loss": 0.3527,
      "step": 1691
    },
    {
      "epoch": 0.62,
      "learning_rate": 3.484121122599705e-05,
      "loss": 0.2963,
      "step": 1692
    },
    {
      "epoch": 0.62,
      "learning_rate": 3.483197932053176e-05,
      "loss": 0.1475,
      "step": 1693
    },
    {
      "epoch": 0.62,
      "learning_rate": 3.482274741506647e-05,
      "loss": 0.21,
      "step": 1694
    },
    {
      "epoch": 0.62,
      "learning_rate": 3.481351550960119e-05,
      "loss": 0.0533,
      "step": 1695
    },
    {
      "epoch": 0.62,
      "learning_rate": 3.48042836041359e-05,
      "loss": 0.1008,
      "step": 1696
    },
    {
      "epoch": 0.62,
      "learning_rate": 3.479505169867061e-05,
      "loss": 0.128,
      "step": 1697
    },
    {
      "epoch": 0.62,
      "learning_rate": 3.478581979320532e-05,
      "loss": 0.2768,
      "step": 1698
    },
    {
      "epoch": 0.62,
      "learning_rate": 3.477658788774003e-05,
      "loss": 0.1154,
      "step": 1699
    },
    {
      "epoch": 0.62,
      "learning_rate": 3.476735598227474e-05,
      "loss": 0.298,
      "step": 1700
    },
    {
      "epoch": 0.62,
      "learning_rate": 3.475812407680946e-05,
      "loss": 0.1748,
      "step": 1701
    },
    {
      "epoch": 0.62,
      "learning_rate": 3.474889217134417e-05,
      "loss": 0.057,
      "step": 1702
    },
    {
      "epoch": 0.62,
      "learning_rate": 3.4739660265878885e-05,
      "loss": 0.0321,
      "step": 1703
    },
    {
      "epoch": 0.62,
      "learning_rate": 3.473042836041359e-05,
      "loss": 0.1312,
      "step": 1704
    },
    {
      "epoch": 0.62,
      "learning_rate": 3.47211964549483e-05,
      "loss": 0.0973,
      "step": 1705
    },
    {
      "epoch": 0.62,
      "learning_rate": 3.4711964549483016e-05,
      "loss": 0.0347,
      "step": 1706
    },
    {
      "epoch": 0.62,
      "learning_rate": 3.470273264401773e-05,
      "loss": 0.0783,
      "step": 1707
    },
    {
      "epoch": 0.62,
      "learning_rate": 3.469350073855244e-05,
      "loss": 0.191,
      "step": 1708
    },
    {
      "epoch": 0.63,
      "learning_rate": 3.4684268833087155e-05,
      "loss": 0.0178,
      "step": 1709
    },
    {
      "epoch": 0.63,
      "learning_rate": 3.4675036927621865e-05,
      "loss": 0.1357,
      "step": 1710
    },
    {
      "epoch": 0.63,
      "learning_rate": 3.466580502215657e-05,
      "loss": 0.011,
      "step": 1711
    },
    {
      "epoch": 0.63,
      "learning_rate": 3.4656573116691286e-05,
      "loss": 0.0046,
      "step": 1712
    },
    {
      "epoch": 0.63,
      "learning_rate": 3.4647341211226e-05,
      "loss": 0.1023,
      "step": 1713
    },
    {
      "epoch": 0.63,
      "learning_rate": 3.4638109305760714e-05,
      "loss": 0.1837,
      "step": 1714
    },
    {
      "epoch": 0.63,
      "learning_rate": 3.4628877400295425e-05,
      "loss": 0.0167,
      "step": 1715
    },
    {
      "epoch": 0.63,
      "learning_rate": 3.4619645494830135e-05,
      "loss": 0.0077,
      "step": 1716
    },
    {
      "epoch": 0.63,
      "learning_rate": 3.4610413589364846e-05,
      "loss": 0.3806,
      "step": 1717
    },
    {
      "epoch": 0.63,
      "learning_rate": 3.4601181683899556e-05,
      "loss": 0.0099,
      "step": 1718
    },
    {
      "epoch": 0.63,
      "learning_rate": 3.459194977843427e-05,
      "loss": 0.2394,
      "step": 1719
    },
    {
      "epoch": 0.63,
      "learning_rate": 3.4582717872968984e-05,
      "loss": 0.1202,
      "step": 1720
    },
    {
      "epoch": 0.63,
      "learning_rate": 3.4573485967503695e-05,
      "loss": 0.0275,
      "step": 1721
    },
    {
      "epoch": 0.63,
      "learning_rate": 3.4564254062038405e-05,
      "loss": 0.0052,
      "step": 1722
    },
    {
      "epoch": 0.63,
      "learning_rate": 3.4555022156573116e-05,
      "loss": 0.0055,
      "step": 1723
    },
    {
      "epoch": 0.63,
      "learning_rate": 3.4545790251107826e-05,
      "loss": 0.0063,
      "step": 1724
    },
    {
      "epoch": 0.63,
      "learning_rate": 3.4536558345642544e-05,
      "loss": 0.0113,
      "step": 1725
    },
    {
      "epoch": 0.63,
      "learning_rate": 3.4527326440177254e-05,
      "loss": 0.0079,
      "step": 1726
    },
    {
      "epoch": 0.63,
      "learning_rate": 3.4518094534711965e-05,
      "loss": 0.2141,
      "step": 1727
    },
    {
      "epoch": 0.63,
      "learning_rate": 3.450886262924668e-05,
      "loss": 0.0093,
      "step": 1728
    },
    {
      "epoch": 0.63,
      "learning_rate": 3.449963072378139e-05,
      "loss": 0.201,
      "step": 1729
    },
    {
      "epoch": 0.63,
      "learning_rate": 3.4490398818316096e-05,
      "loss": 0.0184,
      "step": 1730
    },
    {
      "epoch": 0.63,
      "learning_rate": 3.4481166912850814e-05,
      "loss": 0.391,
      "step": 1731
    },
    {
      "epoch": 0.63,
      "learning_rate": 3.4471935007385524e-05,
      "loss": 0.0066,
      "step": 1732
    },
    {
      "epoch": 0.63,
      "learning_rate": 3.446270310192024e-05,
      "loss": 0.0087,
      "step": 1733
    },
    {
      "epoch": 0.63,
      "learning_rate": 3.445347119645495e-05,
      "loss": 0.0567,
      "step": 1734
    },
    {
      "epoch": 0.63,
      "learning_rate": 3.444423929098966e-05,
      "loss": 0.0363,
      "step": 1735
    },
    {
      "epoch": 0.64,
      "learning_rate": 3.443500738552437e-05,
      "loss": 0.0092,
      "step": 1736
    },
    {
      "epoch": 0.64,
      "learning_rate": 3.4425775480059084e-05,
      "loss": 0.3016,
      "step": 1737
    },
    {
      "epoch": 0.64,
      "learning_rate": 3.4416543574593794e-05,
      "loss": 0.0091,
      "step": 1738
    },
    {
      "epoch": 0.64,
      "learning_rate": 3.440731166912851e-05,
      "loss": 0.0086,
      "step": 1739
    },
    {
      "epoch": 0.64,
      "learning_rate": 3.439807976366322e-05,
      "loss": 0.0048,
      "step": 1740
    },
    {
      "epoch": 0.64,
      "learning_rate": 3.438884785819793e-05,
      "loss": 0.298,
      "step": 1741
    },
    {
      "epoch": 0.64,
      "learning_rate": 3.4379615952732643e-05,
      "loss": 0.0029,
      "step": 1742
    },
    {
      "epoch": 0.64,
      "learning_rate": 3.4370384047267354e-05,
      "loss": 0.1445,
      "step": 1743
    },
    {
      "epoch": 0.64,
      "learning_rate": 3.436115214180207e-05,
      "loss": 0.0053,
      "step": 1744
    },
    {
      "epoch": 0.64,
      "learning_rate": 3.435192023633678e-05,
      "loss": 0.4509,
      "step": 1745
    },
    {
      "epoch": 0.64,
      "learning_rate": 3.434268833087149e-05,
      "loss": 0.0064,
      "step": 1746
    },
    {
      "epoch": 0.64,
      "learning_rate": 3.433345642540621e-05,
      "loss": 0.3118,
      "step": 1747
    },
    {
      "epoch": 0.64,
      "learning_rate": 3.432422451994092e-05,
      "loss": 0.0066,
      "step": 1748
    },
    {
      "epoch": 0.64,
      "learning_rate": 3.4314992614475624e-05,
      "loss": 0.003,
      "step": 1749
    },
    {
      "epoch": 0.64,
      "learning_rate": 3.430576070901034e-05,
      "loss": 0.0043,
      "step": 1750
    },
    {
      "epoch": 0.64,
      "learning_rate": 3.429652880354505e-05,
      "loss": 0.002,
      "step": 1751
    },
    {
      "epoch": 0.64,
      "learning_rate": 3.428729689807976e-05,
      "loss": 0.295,
      "step": 1752
    },
    {
      "epoch": 0.64,
      "learning_rate": 3.427806499261448e-05,
      "loss": 0.0186,
      "step": 1753
    },
    {
      "epoch": 0.64,
      "learning_rate": 3.426883308714919e-05,
      "loss": 0.0015,
      "step": 1754
    },
    {
      "epoch": 0.64,
      "learning_rate": 3.42596011816839e-05,
      "loss": 0.4077,
      "step": 1755
    },
    {
      "epoch": 0.64,
      "learning_rate": 3.425036927621861e-05,
      "loss": 0.0029,
      "step": 1756
    },
    {
      "epoch": 0.64,
      "learning_rate": 3.424113737075332e-05,
      "loss": 0.0033,
      "step": 1757
    },
    {
      "epoch": 0.64,
      "learning_rate": 3.423190546528804e-05,
      "loss": 0.1772,
      "step": 1758
    },
    {
      "epoch": 0.64,
      "learning_rate": 3.422267355982275e-05,
      "loss": 0.0718,
      "step": 1759
    },
    {
      "epoch": 0.64,
      "learning_rate": 3.421344165435746e-05,
      "loss": 0.0227,
      "step": 1760
    },
    {
      "epoch": 0.64,
      "learning_rate": 3.420420974889217e-05,
      "loss": 0.0013,
      "step": 1761
    },
    {
      "epoch": 0.64,
      "learning_rate": 3.419497784342688e-05,
      "loss": 0.0041,
      "step": 1762
    },
    {
      "epoch": 0.65,
      "learning_rate": 3.41857459379616e-05,
      "loss": 0.4827,
      "step": 1763
    },
    {
      "epoch": 0.65,
      "learning_rate": 3.417651403249631e-05,
      "loss": 0.0633,
      "step": 1764
    },
    {
      "epoch": 0.65,
      "learning_rate": 3.416728212703102e-05,
      "loss": 0.0322,
      "step": 1765
    },
    {
      "epoch": 0.65,
      "learning_rate": 3.415805022156574e-05,
      "loss": 0.0025,
      "step": 1766
    },
    {
      "epoch": 0.65,
      "learning_rate": 3.414881831610045e-05,
      "loss": 0.0103,
      "step": 1767
    },
    {
      "epoch": 0.65,
      "learning_rate": 3.413958641063515e-05,
      "loss": 0.0027,
      "step": 1768
    },
    {
      "epoch": 0.65,
      "learning_rate": 3.413035450516987e-05,
      "loss": 0.3567,
      "step": 1769
    },
    {
      "epoch": 0.65,
      "learning_rate": 3.412112259970458e-05,
      "loss": 0.1577,
      "step": 1770
    },
    {
      "epoch": 0.65,
      "learning_rate": 3.411189069423929e-05,
      "loss": 0.2771,
      "step": 1771
    },
    {
      "epoch": 0.65,
      "learning_rate": 3.410265878877401e-05,
      "loss": 0.1278,
      "step": 1772
    },
    {
      "epoch": 0.65,
      "learning_rate": 3.409342688330872e-05,
      "loss": 0.0211,
      "step": 1773
    },
    {
      "epoch": 0.65,
      "learning_rate": 3.408419497784343e-05,
      "loss": 0.0085,
      "step": 1774
    },
    {
      "epoch": 0.65,
      "learning_rate": 3.407496307237814e-05,
      "loss": 0.0042,
      "step": 1775
    },
    {
      "epoch": 0.65,
      "learning_rate": 3.406573116691285e-05,
      "loss": 0.0054,
      "step": 1776
    },
    {
      "epoch": 0.65,
      "learning_rate": 3.405649926144757e-05,
      "loss": 0.0055,
      "step": 1777
    },
    {
      "epoch": 0.65,
      "learning_rate": 3.404726735598228e-05,
      "loss": 0.2211,
      "step": 1778
    },
    {
      "epoch": 0.65,
      "learning_rate": 3.403803545051699e-05,
      "loss": 0.0528,
      "step": 1779
    },
    {
      "epoch": 0.65,
      "learning_rate": 3.40288035450517e-05,
      "loss": 0.2649,
      "step": 1780
    },
    {
      "epoch": 0.65,
      "learning_rate": 3.401957163958641e-05,
      "loss": 0.4068,
      "step": 1781
    },
    {
      "epoch": 0.65,
      "learning_rate": 3.401033973412112e-05,
      "loss": 0.005,
      "step": 1782
    },
    {
      "epoch": 0.65,
      "learning_rate": 3.400110782865584e-05,
      "loss": 0.3123,
      "step": 1783
    },
    {
      "epoch": 0.65,
      "learning_rate": 3.399187592319055e-05,
      "loss": 0.0028,
      "step": 1784
    },
    {
      "epoch": 0.65,
      "learning_rate": 3.3982644017725265e-05,
      "loss": 0.0017,
      "step": 1785
    },
    {
      "epoch": 0.65,
      "learning_rate": 3.3973412112259975e-05,
      "loss": 0.1388,
      "step": 1786
    },
    {
      "epoch": 0.65,
      "learning_rate": 3.396418020679468e-05,
      "loss": 0.0587,
      "step": 1787
    },
    {
      "epoch": 0.65,
      "learning_rate": 3.3954948301329396e-05,
      "loss": 0.0031,
      "step": 1788
    },
    {
      "epoch": 0.65,
      "learning_rate": 3.394571639586411e-05,
      "loss": 0.0056,
      "step": 1789
    },
    {
      "epoch": 0.65,
      "learning_rate": 3.393648449039882e-05,
      "loss": 0.0127,
      "step": 1790
    },
    {
      "epoch": 0.66,
      "learning_rate": 3.3927252584933535e-05,
      "loss": 0.0063,
      "step": 1791
    },
    {
      "epoch": 0.66,
      "learning_rate": 3.3918020679468245e-05,
      "loss": 0.0905,
      "step": 1792
    },
    {
      "epoch": 0.66,
      "learning_rate": 3.3908788774002956e-05,
      "loss": 0.0035,
      "step": 1793
    },
    {
      "epoch": 0.66,
      "learning_rate": 3.3899556868537666e-05,
      "loss": 0.2347,
      "step": 1794
    },
    {
      "epoch": 0.66,
      "learning_rate": 3.389032496307238e-05,
      "loss": 0.008,
      "step": 1795
    },
    {
      "epoch": 0.66,
      "learning_rate": 3.3881093057607094e-05,
      "loss": 0.0904,
      "step": 1796
    },
    {
      "epoch": 0.66,
      "learning_rate": 3.3871861152141805e-05,
      "loss": 0.2265,
      "step": 1797
    },
    {
      "epoch": 0.66,
      "learning_rate": 3.3862629246676515e-05,
      "loss": 0.0215,
      "step": 1798
    },
    {
      "epoch": 0.66,
      "learning_rate": 3.3853397341211226e-05,
      "loss": 0.027,
      "step": 1799
    },
    {
      "epoch": 0.66,
      "learning_rate": 3.3844165435745937e-05,
      "loss": 0.0011,
      "step": 1800
    },
    {
      "epoch": 0.66,
      "learning_rate": 3.383493353028065e-05,
      "loss": 0.4109,
      "step": 1801
    },
    {
      "epoch": 0.66,
      "learning_rate": 3.3825701624815364e-05,
      "loss": 0.7002,
      "step": 1802
    },
    {
      "epoch": 0.66,
      "learning_rate": 3.3816469719350075e-05,
      "loss": 0.004,
      "step": 1803
    },
    {
      "epoch": 0.66,
      "learning_rate": 3.380723781388479e-05,
      "loss": 0.0019,
      "step": 1804
    },
    {
      "epoch": 0.66,
      "learning_rate": 3.37980059084195e-05,
      "loss": 0.0041,
      "step": 1805
    },
    {
      "epoch": 0.66,
      "learning_rate": 3.378877400295421e-05,
      "loss": 0.4833,
      "step": 1806
    },
    {
      "epoch": 0.66,
      "learning_rate": 3.3779542097488924e-05,
      "loss": 0.0111,
      "step": 1807
    },
    {
      "epoch": 0.66,
      "learning_rate": 3.3770310192023635e-05,
      "loss": 0.0131,
      "step": 1808
    },
    {
      "epoch": 0.66,
      "learning_rate": 3.3761078286558345e-05,
      "loss": 0.01,
      "step": 1809
    },
    {
      "epoch": 0.66,
      "learning_rate": 3.375184638109306e-05,
      "loss": 0.0055,
      "step": 1810
    },
    {
      "epoch": 0.66,
      "learning_rate": 3.374261447562777e-05,
      "loss": 0.0045,
      "step": 1811
    },
    {
      "epoch": 0.66,
      "learning_rate": 3.3733382570162483e-05,
      "loss": 0.3128,
      "step": 1812
    },
    {
      "epoch": 0.66,
      "learning_rate": 3.3724150664697194e-05,
      "loss": 0.2294,
      "step": 1813
    },
    {
      "epoch": 0.66,
      "learning_rate": 3.3714918759231905e-05,
      "loss": 0.0105,
      "step": 1814
    },
    {
      "epoch": 0.66,
      "learning_rate": 3.370568685376662e-05,
      "loss": 0.0026,
      "step": 1815
    },
    {
      "epoch": 0.66,
      "learning_rate": 3.369645494830133e-05,
      "loss": 0.005,
      "step": 1816
    },
    {
      "epoch": 0.66,
      "learning_rate": 3.368722304283604e-05,
      "loss": 0.295,
      "step": 1817
    },
    {
      "epoch": 0.67,
      "learning_rate": 3.3677991137370754e-05,
      "loss": 0.0057,
      "step": 1818
    },
    {
      "epoch": 0.67,
      "learning_rate": 3.3668759231905464e-05,
      "loss": 0.0026,
      "step": 1819
    },
    {
      "epoch": 0.67,
      "learning_rate": 3.3659527326440175e-05,
      "loss": 0.1492,
      "step": 1820
    },
    {
      "epoch": 0.67,
      "learning_rate": 3.365029542097489e-05,
      "loss": 0.0139,
      "step": 1821
    },
    {
      "epoch": 0.67,
      "learning_rate": 3.36410635155096e-05,
      "loss": 0.2042,
      "step": 1822
    },
    {
      "epoch": 0.67,
      "learning_rate": 3.363183161004431e-05,
      "loss": 0.0021,
      "step": 1823
    },
    {
      "epoch": 0.67,
      "learning_rate": 3.362259970457903e-05,
      "loss": 0.2153,
      "step": 1824
    },
    {
      "epoch": 0.67,
      "learning_rate": 3.3613367799113734e-05,
      "loss": 0.0052,
      "step": 1825
    },
    {
      "epoch": 0.67,
      "learning_rate": 3.360413589364845e-05,
      "loss": 0.0143,
      "step": 1826
    },
    {
      "epoch": 0.67,
      "learning_rate": 3.359490398818316e-05,
      "loss": 0.0042,
      "step": 1827
    },
    {
      "epoch": 0.67,
      "learning_rate": 3.358567208271787e-05,
      "loss": 0.0124,
      "step": 1828
    },
    {
      "epoch": 0.67,
      "learning_rate": 3.357644017725259e-05,
      "loss": 0.0042,
      "step": 1829
    },
    {
      "epoch": 0.67,
      "learning_rate": 3.35672082717873e-05,
      "loss": 0.1259,
      "step": 1830
    },
    {
      "epoch": 0.67,
      "learning_rate": 3.355797636632201e-05,
      "loss": 0.0416,
      "step": 1831
    },
    {
      "epoch": 0.67,
      "learning_rate": 3.354874446085672e-05,
      "loss": 0.47,
      "step": 1832
    },
    {
      "epoch": 0.67,
      "learning_rate": 3.353951255539143e-05,
      "loss": 0.0996,
      "step": 1833
    },
    {
      "epoch": 0.67,
      "learning_rate": 3.353028064992615e-05,
      "loss": 0.2198,
      "step": 1834
    },
    {
      "epoch": 0.67,
      "learning_rate": 3.352104874446086e-05,
      "loss": 0.4997,
      "step": 1835
    },
    {
      "epoch": 0.67,
      "learning_rate": 3.351181683899557e-05,
      "loss": 0.0617,
      "step": 1836
    },
    {
      "epoch": 0.67,
      "learning_rate": 3.350258493353028e-05,
      "loss": 0.2387,
      "step": 1837
    },
    {
      "epoch": 0.67,
      "learning_rate": 3.349335302806499e-05,
      "loss": 0.0055,
      "step": 1838
    },
    {
      "epoch": 0.67,
      "learning_rate": 3.34841211225997e-05,
      "loss": 0.2837,
      "step": 1839
    },
    {
      "epoch": 0.67,
      "learning_rate": 3.347488921713442e-05,
      "loss": 0.0173,
      "step": 1840
    },
    {
      "epoch": 0.67,
      "learning_rate": 3.346565731166913e-05,
      "loss": 0.0038,
      "step": 1841
    },
    {
      "epoch": 0.67,
      "learning_rate": 3.345642540620384e-05,
      "loss": 0.0645,
      "step": 1842
    },
    {
      "epoch": 0.67,
      "learning_rate": 3.344719350073856e-05,
      "loss": 0.0653,
      "step": 1843
    },
    {
      "epoch": 0.67,
      "learning_rate": 3.343796159527326e-05,
      "loss": 0.3303,
      "step": 1844
    },
    {
      "epoch": 0.68,
      "learning_rate": 3.342872968980798e-05,
      "loss": 0.2031,
      "step": 1845
    },
    {
      "epoch": 0.68,
      "learning_rate": 3.341949778434269e-05,
      "loss": 0.2433,
      "step": 1846
    },
    {
      "epoch": 0.68,
      "learning_rate": 3.34102658788774e-05,
      "loss": 0.7227,
      "step": 1847
    },
    {
      "epoch": 0.68,
      "learning_rate": 3.340103397341212e-05,
      "loss": 0.0398,
      "step": 1848
    },
    {
      "epoch": 0.68,
      "learning_rate": 3.339180206794683e-05,
      "loss": 0.1296,
      "step": 1849
    },
    {
      "epoch": 0.68,
      "learning_rate": 3.338257016248154e-05,
      "loss": 0.202,
      "step": 1850
    },
    {
      "epoch": 0.68,
      "learning_rate": 3.337333825701625e-05,
      "loss": 0.0212,
      "step": 1851
    },
    {
      "epoch": 0.68,
      "learning_rate": 3.336410635155096e-05,
      "loss": 0.0994,
      "step": 1852
    },
    {
      "epoch": 0.68,
      "learning_rate": 3.335487444608567e-05,
      "loss": 0.1701,
      "step": 1853
    },
    {
      "epoch": 0.68,
      "learning_rate": 3.334564254062039e-05,
      "loss": 0.0115,
      "step": 1854
    },
    {
      "epoch": 0.68,
      "learning_rate": 3.33364106351551e-05,
      "loss": 0.2356,
      "step": 1855
    },
    {
      "epoch": 0.68,
      "learning_rate": 3.332717872968981e-05,
      "loss": 0.0192,
      "step": 1856
    },
    {
      "epoch": 0.68,
      "learning_rate": 3.331794682422452e-05,
      "loss": 0.0055,
      "step": 1857
    },
    {
      "epoch": 0.68,
      "learning_rate": 3.330871491875923e-05,
      "loss": 0.0183,
      "step": 1858
    },
    {
      "epoch": 0.68,
      "learning_rate": 3.329948301329395e-05,
      "loss": 0.2746,
      "step": 1859
    },
    {
      "epoch": 0.68,
      "learning_rate": 3.329025110782866e-05,
      "loss": 0.3399,
      "step": 1860
    },
    {
      "epoch": 0.68,
      "learning_rate": 3.328101920236337e-05,
      "loss": 0.0146,
      "step": 1861
    },
    {
      "epoch": 0.68,
      "learning_rate": 3.3271787296898085e-05,
      "loss": 0.0293,
      "step": 1862
    },
    {
      "epoch": 0.68,
      "learning_rate": 3.326255539143279e-05,
      "loss": 0.0118,
      "step": 1863
    },
    {
      "epoch": 0.68,
      "learning_rate": 3.3253323485967507e-05,
      "loss": 0.0188,
      "step": 1864
    },
    {
      "epoch": 0.68,
      "learning_rate": 3.324409158050222e-05,
      "loss": 0.0035,
      "step": 1865
    },
    {
      "epoch": 0.68,
      "learning_rate": 3.323485967503693e-05,
      "loss": 0.2275,
      "step": 1866
    },
    {
      "epoch": 0.68,
      "learning_rate": 3.3225627769571645e-05,
      "loss": 0.0069,
      "step": 1867
    },
    {
      "epoch": 0.68,
      "learning_rate": 3.3216395864106356e-05,
      "loss": 0.4009,
      "step": 1868
    },
    {
      "epoch": 0.68,
      "learning_rate": 3.3207163958641066e-05,
      "loss": 0.003,
      "step": 1869
    },
    {
      "epoch": 0.68,
      "learning_rate": 3.319793205317578e-05,
      "loss": 0.3068,
      "step": 1870
    },
    {
      "epoch": 0.68,
      "learning_rate": 3.318870014771049e-05,
      "loss": 0.0103,
      "step": 1871
    },
    {
      "epoch": 0.68,
      "learning_rate": 3.31794682422452e-05,
      "loss": 0.0034,
      "step": 1872
    },
    {
      "epoch": 0.69,
      "learning_rate": 3.3170236336779915e-05,
      "loss": 0.0037,
      "step": 1873
    },
    {
      "epoch": 0.69,
      "learning_rate": 3.3161004431314626e-05,
      "loss": 0.5512,
      "step": 1874
    },
    {
      "epoch": 0.69,
      "learning_rate": 3.3151772525849336e-05,
      "loss": 0.3409,
      "step": 1875
    },
    {
      "epoch": 0.69,
      "learning_rate": 3.314254062038405e-05,
      "loss": 0.0051,
      "step": 1876
    },
    {
      "epoch": 0.69,
      "learning_rate": 3.313330871491876e-05,
      "loss": 0.0049,
      "step": 1877
    },
    {
      "epoch": 0.69,
      "learning_rate": 3.3124076809453475e-05,
      "loss": 0.0819,
      "step": 1878
    },
    {
      "epoch": 0.69,
      "learning_rate": 3.3114844903988185e-05,
      "loss": 0.0128,
      "step": 1879
    },
    {
      "epoch": 0.69,
      "learning_rate": 3.3105612998522896e-05,
      "loss": 0.1641,
      "step": 1880
    },
    {
      "epoch": 0.69,
      "learning_rate": 3.309638109305761e-05,
      "loss": 0.1562,
      "step": 1881
    },
    {
      "epoch": 0.69,
      "learning_rate": 3.308714918759232e-05,
      "loss": 0.327,
      "step": 1882
    },
    {
      "epoch": 0.69,
      "learning_rate": 3.307791728212703e-05,
      "loss": 0.0041,
      "step": 1883
    },
    {
      "epoch": 0.69,
      "learning_rate": 3.3068685376661745e-05,
      "loss": 0.0072,
      "step": 1884
    },
    {
      "epoch": 0.69,
      "learning_rate": 3.3059453471196455e-05,
      "loss": 0.002,
      "step": 1885
    },
    {
      "epoch": 0.69,
      "learning_rate": 3.305022156573117e-05,
      "loss": 0.0122,
      "step": 1886
    },
    {
      "epoch": 0.69,
      "learning_rate": 3.304098966026588e-05,
      "loss": 0.0061,
      "step": 1887
    },
    {
      "epoch": 0.69,
      "learning_rate": 3.3031757754800594e-05,
      "loss": 0.0076,
      "step": 1888
    },
    {
      "epoch": 0.69,
      "learning_rate": 3.3022525849335304e-05,
      "loss": 0.0077,
      "step": 1889
    },
    {
      "epoch": 0.69,
      "learning_rate": 3.3013293943870015e-05,
      "loss": 0.0035,
      "step": 1890
    },
    {
      "epoch": 0.69,
      "learning_rate": 3.3004062038404725e-05,
      "loss": 0.0104,
      "step": 1891
    },
    {
      "epoch": 0.69,
      "learning_rate": 3.299483013293944e-05,
      "loss": 0.2089,
      "step": 1892
    },
    {
      "epoch": 0.69,
      "learning_rate": 3.298559822747415e-05,
      "loss": 0.3211,
      "step": 1893
    },
    {
      "epoch": 0.69,
      "learning_rate": 3.2976366322008864e-05,
      "loss": 0.0994,
      "step": 1894
    },
    {
      "epoch": 0.69,
      "learning_rate": 3.2967134416543574e-05,
      "loss": 0.0078,
      "step": 1895
    },
    {
      "epoch": 0.69,
      "learning_rate": 3.2957902511078285e-05,
      "loss": 0.0082,
      "step": 1896
    },
    {
      "epoch": 0.69,
      "learning_rate": 3.2948670605613e-05,
      "loss": 0.7891,
      "step": 1897
    },
    {
      "epoch": 0.69,
      "learning_rate": 3.293943870014771e-05,
      "loss": 0.0051,
      "step": 1898
    },
    {
      "epoch": 0.69,
      "learning_rate": 3.293020679468242e-05,
      "loss": 0.0027,
      "step": 1899
    },
    {
      "epoch": 0.7,
      "learning_rate": 3.292097488921714e-05,
      "loss": 0.0443,
      "step": 1900
    },
    {
      "epoch": 0.7,
      "learning_rate": 3.2911742983751844e-05,
      "loss": 0.3649,
      "step": 1901
    },
    {
      "epoch": 0.7,
      "learning_rate": 3.2902511078286555e-05,
      "loss": 0.0061,
      "step": 1902
    },
    {
      "epoch": 0.7,
      "learning_rate": 3.289327917282127e-05,
      "loss": 0.0086,
      "step": 1903
    },
    {
      "epoch": 0.7,
      "learning_rate": 3.288404726735598e-05,
      "loss": 0.1254,
      "step": 1904
    },
    {
      "epoch": 0.7,
      "learning_rate": 3.28748153618907e-05,
      "loss": 0.3495,
      "step": 1905
    },
    {
      "epoch": 0.7,
      "learning_rate": 3.286558345642541e-05,
      "loss": 0.0225,
      "step": 1906
    },
    {
      "epoch": 0.7,
      "learning_rate": 3.285635155096012e-05,
      "loss": 0.3974,
      "step": 1907
    },
    {
      "epoch": 0.7,
      "learning_rate": 3.284711964549483e-05,
      "loss": 0.1425,
      "step": 1908
    },
    {
      "epoch": 0.7,
      "learning_rate": 3.283788774002954e-05,
      "loss": 0.5126,
      "step": 1909
    },
    {
      "epoch": 0.7,
      "learning_rate": 3.282865583456425e-05,
      "loss": 0.0028,
      "step": 1910
    },
    {
      "epoch": 0.7,
      "learning_rate": 3.281942392909897e-05,
      "loss": 0.2607,
      "step": 1911
    },
    {
      "epoch": 0.7,
      "learning_rate": 3.281019202363368e-05,
      "loss": 0.1871,
      "step": 1912
    },
    {
      "epoch": 0.7,
      "learning_rate": 3.280096011816839e-05,
      "loss": 0.0089,
      "step": 1913
    },
    {
      "epoch": 0.7,
      "learning_rate": 3.27917282127031e-05,
      "loss": 0.0165,
      "step": 1914
    },
    {
      "epoch": 0.7,
      "learning_rate": 3.278249630723781e-05,
      "loss": 0.108,
      "step": 1915
    },
    {
      "epoch": 0.7,
      "learning_rate": 3.277326440177253e-05,
      "loss": 0.3283,
      "step": 1916
    },
    {
      "epoch": 0.7,
      "learning_rate": 3.276403249630724e-05,
      "loss": 0.2172,
      "step": 1917
    },
    {
      "epoch": 0.7,
      "learning_rate": 3.275480059084195e-05,
      "loss": 0.2825,
      "step": 1918
    },
    {
      "epoch": 0.7,
      "learning_rate": 3.274556868537667e-05,
      "loss": 0.0049,
      "step": 1919
    },
    {
      "epoch": 0.7,
      "learning_rate": 3.273633677991137e-05,
      "loss": 0.1274,
      "step": 1920
    },
    {
      "epoch": 0.7,
      "learning_rate": 3.272710487444608e-05,
      "loss": 0.0097,
      "step": 1921
    },
    {
      "epoch": 0.7,
      "learning_rate": 3.27178729689808e-05,
      "loss": 0.0106,
      "step": 1922
    },
    {
      "epoch": 0.7,
      "learning_rate": 3.270864106351551e-05,
      "loss": 0.1588,
      "step": 1923
    },
    {
      "epoch": 0.7,
      "learning_rate": 3.269940915805022e-05,
      "loss": 0.3397,
      "step": 1924
    },
    {
      "epoch": 0.7,
      "learning_rate": 3.269017725258494e-05,
      "loss": 0.0139,
      "step": 1925
    },
    {
      "epoch": 0.7,
      "learning_rate": 3.268094534711965e-05,
      "loss": 0.0297,
      "step": 1926
    },
    {
      "epoch": 0.71,
      "learning_rate": 3.267171344165436e-05,
      "loss": 0.0416,
      "step": 1927
    },
    {
      "epoch": 0.71,
      "learning_rate": 3.266248153618907e-05,
      "loss": 0.1106,
      "step": 1928
    },
    {
      "epoch": 0.71,
      "learning_rate": 3.265324963072378e-05,
      "loss": 0.0224,
      "step": 1929
    },
    {
      "epoch": 0.71,
      "learning_rate": 3.26440177252585e-05,
      "loss": 0.0124,
      "step": 1930
    },
    {
      "epoch": 0.71,
      "learning_rate": 3.263478581979321e-05,
      "loss": 0.0097,
      "step": 1931
    },
    {
      "epoch": 0.71,
      "learning_rate": 3.262555391432792e-05,
      "loss": 0.0332,
      "step": 1932
    },
    {
      "epoch": 0.71,
      "learning_rate": 3.261632200886263e-05,
      "loss": 0.0919,
      "step": 1933
    },
    {
      "epoch": 0.71,
      "learning_rate": 3.260709010339734e-05,
      "loss": 0.0097,
      "step": 1934
    },
    {
      "epoch": 0.71,
      "learning_rate": 3.259785819793206e-05,
      "loss": 0.0199,
      "step": 1935
    },
    {
      "epoch": 0.71,
      "learning_rate": 3.258862629246677e-05,
      "loss": 0.2105,
      "step": 1936
    },
    {
      "epoch": 0.71,
      "learning_rate": 3.257939438700148e-05,
      "loss": 0.0057,
      "step": 1937
    },
    {
      "epoch": 0.71,
      "learning_rate": 3.2570162481536196e-05,
      "loss": 0.0017,
      "step": 1938
    },
    {
      "epoch": 0.71,
      "learning_rate": 3.25609305760709e-05,
      "loss": 0.0522,
      "step": 1939
    },
    {
      "epoch": 0.71,
      "learning_rate": 3.255169867060561e-05,
      "loss": 0.0202,
      "step": 1940
    },
    {
      "epoch": 0.71,
      "learning_rate": 3.254246676514033e-05,
      "loss": 0.0062,
      "step": 1941
    },
    {
      "epoch": 0.71,
      "learning_rate": 3.253323485967504e-05,
      "loss": 0.7772,
      "step": 1942
    },
    {
      "epoch": 0.71,
      "learning_rate": 3.252400295420975e-05,
      "loss": 0.2516,
      "step": 1943
    },
    {
      "epoch": 0.71,
      "learning_rate": 3.2514771048744466e-05,
      "loss": 0.0031,
      "step": 1944
    },
    {
      "epoch": 0.71,
      "learning_rate": 3.2505539143279176e-05,
      "loss": 0.0243,
      "step": 1945
    },
    {
      "epoch": 0.71,
      "learning_rate": 3.249630723781389e-05,
      "loss": 0.1698,
      "step": 1946
    },
    {
      "epoch": 0.71,
      "learning_rate": 3.24870753323486e-05,
      "loss": 0.003,
      "step": 1947
    },
    {
      "epoch": 0.71,
      "learning_rate": 3.247784342688331e-05,
      "loss": 0.0031,
      "step": 1948
    },
    {
      "epoch": 0.71,
      "learning_rate": 3.2468611521418025e-05,
      "loss": 0.0093,
      "step": 1949
    },
    {
      "epoch": 0.71,
      "learning_rate": 3.2459379615952736e-05,
      "loss": 0.092,
      "step": 1950
    },
    {
      "epoch": 0.71,
      "learning_rate": 3.2450147710487446e-05,
      "loss": 0.1637,
      "step": 1951
    },
    {
      "epoch": 0.71,
      "learning_rate": 3.244091580502216e-05,
      "loss": 0.0044,
      "step": 1952
    },
    {
      "epoch": 0.71,
      "learning_rate": 3.243168389955687e-05,
      "loss": 0.0559,
      "step": 1953
    },
    {
      "epoch": 0.71,
      "learning_rate": 3.242245199409158e-05,
      "loss": 0.0025,
      "step": 1954
    },
    {
      "epoch": 0.72,
      "learning_rate": 3.2413220088626295e-05,
      "loss": 0.291,
      "step": 1955
    },
    {
      "epoch": 0.72,
      "learning_rate": 3.2403988183161006e-05,
      "loss": 0.0048,
      "step": 1956
    },
    {
      "epoch": 0.72,
      "learning_rate": 3.239475627769572e-05,
      "loss": 0.3104,
      "step": 1957
    },
    {
      "epoch": 0.72,
      "learning_rate": 3.238552437223043e-05,
      "loss": 0.0126,
      "step": 1958
    },
    {
      "epoch": 0.72,
      "learning_rate": 3.237629246676514e-05,
      "loss": 0.0032,
      "step": 1959
    },
    {
      "epoch": 0.72,
      "learning_rate": 3.2367060561299855e-05,
      "loss": 0.015,
      "step": 1960
    },
    {
      "epoch": 0.72,
      "learning_rate": 3.2357828655834565e-05,
      "loss": 0.1675,
      "step": 1961
    },
    {
      "epoch": 0.72,
      "learning_rate": 3.2348596750369276e-05,
      "loss": 0.0037,
      "step": 1962
    },
    {
      "epoch": 0.72,
      "learning_rate": 3.233936484490399e-05,
      "loss": 0.4256,
      "step": 1963
    },
    {
      "epoch": 0.72,
      "learning_rate": 3.2330132939438704e-05,
      "loss": 0.0039,
      "step": 1964
    },
    {
      "epoch": 0.72,
      "learning_rate": 3.2320901033973414e-05,
      "loss": 0.0056,
      "step": 1965
    },
    {
      "epoch": 0.72,
      "learning_rate": 3.2311669128508125e-05,
      "loss": 0.3681,
      "step": 1966
    },
    {
      "epoch": 0.72,
      "learning_rate": 3.2302437223042835e-05,
      "loss": 0.0077,
      "step": 1967
    },
    {
      "epoch": 0.72,
      "learning_rate": 3.229320531757755e-05,
      "loss": 0.0108,
      "step": 1968
    },
    {
      "epoch": 0.72,
      "learning_rate": 3.228397341211226e-05,
      "loss": 0.0054,
      "step": 1969
    },
    {
      "epoch": 0.72,
      "learning_rate": 3.2274741506646974e-05,
      "loss": 0.0089,
      "step": 1970
    },
    {
      "epoch": 0.72,
      "learning_rate": 3.2265509601181684e-05,
      "loss": 0.0181,
      "step": 1971
    },
    {
      "epoch": 0.72,
      "learning_rate": 3.2256277695716395e-05,
      "loss": 0.3907,
      "step": 1972
    },
    {
      "epoch": 0.72,
      "learning_rate": 3.2247045790251105e-05,
      "loss": 0.0024,
      "step": 1973
    },
    {
      "epoch": 0.72,
      "learning_rate": 3.223781388478582e-05,
      "loss": 0.3024,
      "step": 1974
    },
    {
      "epoch": 0.72,
      "learning_rate": 3.222858197932053e-05,
      "loss": 0.3732,
      "step": 1975
    },
    {
      "epoch": 0.72,
      "learning_rate": 3.221935007385525e-05,
      "loss": 0.3233,
      "step": 1976
    },
    {
      "epoch": 0.72,
      "learning_rate": 3.2210118168389954e-05,
      "loss": 0.4112,
      "step": 1977
    },
    {
      "epoch": 0.72,
      "learning_rate": 3.2200886262924665e-05,
      "loss": 0.0042,
      "step": 1978
    },
    {
      "epoch": 0.72,
      "learning_rate": 3.219165435745938e-05,
      "loss": 0.0082,
      "step": 1979
    },
    {
      "epoch": 0.72,
      "learning_rate": 3.218242245199409e-05,
      "loss": 0.2112,
      "step": 1980
    },
    {
      "epoch": 0.72,
      "learning_rate": 3.2173190546528803e-05,
      "loss": 0.0218,
      "step": 1981
    },
    {
      "epoch": 0.73,
      "learning_rate": 3.216395864106352e-05,
      "loss": 0.0047,
      "step": 1982
    },
    {
      "epoch": 0.73,
      "learning_rate": 3.215472673559823e-05,
      "loss": 0.1811,
      "step": 1983
    },
    {
      "epoch": 0.73,
      "learning_rate": 3.2145494830132935e-05,
      "loss": 0.4397,
      "step": 1984
    },
    {
      "epoch": 0.73,
      "learning_rate": 3.213626292466765e-05,
      "loss": 0.0052,
      "step": 1985
    },
    {
      "epoch": 0.73,
      "learning_rate": 3.212703101920236e-05,
      "loss": 0.0123,
      "step": 1986
    },
    {
      "epoch": 0.73,
      "learning_rate": 3.211779911373708e-05,
      "loss": 0.0138,
      "step": 1987
    },
    {
      "epoch": 0.73,
      "learning_rate": 3.210856720827179e-05,
      "loss": 0.009,
      "step": 1988
    },
    {
      "epoch": 0.73,
      "learning_rate": 3.20993353028065e-05,
      "loss": 0.0071,
      "step": 1989
    },
    {
      "epoch": 0.73,
      "learning_rate": 3.209010339734121e-05,
      "loss": 0.087,
      "step": 1990
    },
    {
      "epoch": 0.73,
      "learning_rate": 3.208087149187592e-05,
      "loss": 0.0457,
      "step": 1991
    },
    {
      "epoch": 0.73,
      "learning_rate": 3.207163958641063e-05,
      "loss": 0.3005,
      "step": 1992
    },
    {
      "epoch": 0.73,
      "learning_rate": 3.206240768094535e-05,
      "loss": 0.6813,
      "step": 1993
    },
    {
      "epoch": 0.73,
      "learning_rate": 3.205317577548006e-05,
      "loss": 0.38,
      "step": 1994
    },
    {
      "epoch": 0.73,
      "learning_rate": 3.204394387001477e-05,
      "loss": 0.4935,
      "step": 1995
    },
    {
      "epoch": 0.73,
      "learning_rate": 3.203471196454948e-05,
      "loss": 0.2417,
      "step": 1996
    },
    {
      "epoch": 0.73,
      "learning_rate": 3.202548005908419e-05,
      "loss": 0.3926,
      "step": 1997
    },
    {
      "epoch": 0.73,
      "learning_rate": 3.201624815361891e-05,
      "loss": 0.0269,
      "step": 1998
    },
    {
      "epoch": 0.73,
      "learning_rate": 3.200701624815362e-05,
      "loss": 0.0086,
      "step": 1999
    },
    {
      "epoch": 0.73,
      "learning_rate": 3.199778434268833e-05,
      "loss": 0.179,
      "step": 2000
    },
    {
      "epoch": 0.73,
      "learning_rate": 3.198855243722305e-05,
      "loss": 0.0357,
      "step": 2001
    },
    {
      "epoch": 0.73,
      "learning_rate": 3.197932053175776e-05,
      "loss": 0.147,
      "step": 2002
    },
    {
      "epoch": 0.73,
      "learning_rate": 3.197008862629246e-05,
      "loss": 0.0114,
      "step": 2003
    },
    {
      "epoch": 0.73,
      "learning_rate": 3.196085672082718e-05,
      "loss": 0.2549,
      "step": 2004
    },
    {
      "epoch": 0.73,
      "learning_rate": 3.195162481536189e-05,
      "loss": 0.2388,
      "step": 2005
    },
    {
      "epoch": 0.73,
      "learning_rate": 3.194239290989661e-05,
      "loss": 0.0069,
      "step": 2006
    },
    {
      "epoch": 0.73,
      "learning_rate": 3.193316100443132e-05,
      "loss": 0.0126,
      "step": 2007
    },
    {
      "epoch": 0.73,
      "learning_rate": 3.192392909896603e-05,
      "loss": 0.19,
      "step": 2008
    },
    {
      "epoch": 0.74,
      "learning_rate": 3.191469719350074e-05,
      "loss": 0.0131,
      "step": 2009
    },
    {
      "epoch": 0.74,
      "learning_rate": 3.190546528803545e-05,
      "loss": 0.0075,
      "step": 2010
    },
    {
      "epoch": 0.74,
      "learning_rate": 3.189623338257016e-05,
      "loss": 0.2527,
      "step": 2011
    },
    {
      "epoch": 0.74,
      "learning_rate": 3.188700147710488e-05,
      "loss": 0.0244,
      "step": 2012
    },
    {
      "epoch": 0.74,
      "learning_rate": 3.187776957163959e-05,
      "loss": 0.7417,
      "step": 2013
    },
    {
      "epoch": 0.74,
      "learning_rate": 3.18685376661743e-05,
      "loss": 0.2802,
      "step": 2014
    },
    {
      "epoch": 0.74,
      "learning_rate": 3.185930576070901e-05,
      "loss": 0.2688,
      "step": 2015
    },
    {
      "epoch": 0.74,
      "learning_rate": 3.185007385524372e-05,
      "loss": 0.2637,
      "step": 2016
    },
    {
      "epoch": 0.74,
      "learning_rate": 3.184084194977844e-05,
      "loss": 0.014,
      "step": 2017
    },
    {
      "epoch": 0.74,
      "learning_rate": 3.183161004431315e-05,
      "loss": 0.0291,
      "step": 2018
    },
    {
      "epoch": 0.74,
      "learning_rate": 3.182237813884786e-05,
      "loss": 0.0221,
      "step": 2019
    },
    {
      "epoch": 0.74,
      "learning_rate": 3.1813146233382576e-05,
      "loss": 0.179,
      "step": 2020
    },
    {
      "epoch": 0.74,
      "learning_rate": 3.1803914327917286e-05,
      "loss": 0.0305,
      "step": 2021
    },
    {
      "epoch": 0.74,
      "learning_rate": 3.179468242245199e-05,
      "loss": 0.0171,
      "step": 2022
    },
    {
      "epoch": 0.74,
      "learning_rate": 3.178545051698671e-05,
      "loss": 0.0106,
      "step": 2023
    },
    {
      "epoch": 0.74,
      "learning_rate": 3.177621861152142e-05,
      "loss": 0.0422,
      "step": 2024
    },
    {
      "epoch": 0.74,
      "learning_rate": 3.176698670605613e-05,
      "loss": 0.0167,
      "step": 2025
    },
    {
      "epoch": 0.74,
      "learning_rate": 3.1757754800590846e-05,
      "loss": 0.0129,
      "step": 2026
    },
    {
      "epoch": 0.74,
      "learning_rate": 3.1748522895125556e-05,
      "loss": 0.0312,
      "step": 2027
    },
    {
      "epoch": 0.74,
      "learning_rate": 3.173929098966027e-05,
      "loss": 0.0197,
      "step": 2028
    },
    {
      "epoch": 0.74,
      "learning_rate": 3.173005908419498e-05,
      "loss": 0.0233,
      "step": 2029
    },
    {
      "epoch": 0.74,
      "learning_rate": 3.172082717872969e-05,
      "loss": 0.0125,
      "step": 2030
    },
    {
      "epoch": 0.74,
      "learning_rate": 3.1711595273264405e-05,
      "loss": 0.0314,
      "step": 2031
    },
    {
      "epoch": 0.74,
      "learning_rate": 3.1702363367799116e-05,
      "loss": 0.119,
      "step": 2032
    },
    {
      "epoch": 0.74,
      "learning_rate": 3.1693131462333827e-05,
      "loss": 0.0047,
      "step": 2033
    },
    {
      "epoch": 0.74,
      "learning_rate": 3.168389955686854e-05,
      "loss": 0.0644,
      "step": 2034
    },
    {
      "epoch": 0.74,
      "learning_rate": 3.167466765140325e-05,
      "loss": 0.2412,
      "step": 2035
    },
    {
      "epoch": 0.74,
      "learning_rate": 3.1665435745937965e-05,
      "loss": 0.0125,
      "step": 2036
    },
    {
      "epoch": 0.75,
      "learning_rate": 3.1656203840472676e-05,
      "loss": 0.0085,
      "step": 2037
    },
    {
      "epoch": 0.75,
      "learning_rate": 3.1646971935007386e-05,
      "loss": 0.2849,
      "step": 2038
    },
    {
      "epoch": 0.75,
      "learning_rate": 3.16377400295421e-05,
      "loss": 0.0655,
      "step": 2039
    },
    {
      "epoch": 0.75,
      "learning_rate": 3.1628508124076814e-05,
      "loss": 0.3465,
      "step": 2040
    },
    {
      "epoch": 0.75,
      "learning_rate": 3.161927621861152e-05,
      "loss": 0.0056,
      "step": 2041
    },
    {
      "epoch": 0.75,
      "learning_rate": 3.1610044313146235e-05,
      "loss": 0.1168,
      "step": 2042
    },
    {
      "epoch": 0.75,
      "learning_rate": 3.1600812407680946e-05,
      "loss": 0.0018,
      "step": 2043
    },
    {
      "epoch": 0.75,
      "learning_rate": 3.1591580502215656e-05,
      "loss": 0.009,
      "step": 2044
    },
    {
      "epoch": 0.75,
      "learning_rate": 3.1582348596750373e-05,
      "loss": 0.2353,
      "step": 2045
    },
    {
      "epoch": 0.75,
      "learning_rate": 3.1573116691285084e-05,
      "loss": 0.605,
      "step": 2046
    },
    {
      "epoch": 0.75,
      "learning_rate": 3.1563884785819795e-05,
      "loss": 0.0122,
      "step": 2047
    },
    {
      "epoch": 0.75,
      "learning_rate": 3.1554652880354505e-05,
      "loss": 0.2122,
      "step": 2048
    },
    {
      "epoch": 0.75,
      "learning_rate": 3.1545420974889216e-05,
      "loss": 0.0041,
      "step": 2049
    },
    {
      "epoch": 0.75,
      "learning_rate": 3.153618906942393e-05,
      "loss": 0.0038,
      "step": 2050
    },
    {
      "epoch": 0.75,
      "learning_rate": 3.1526957163958644e-05,
      "loss": 0.0064,
      "step": 2051
    },
    {
      "epoch": 0.75,
      "learning_rate": 3.1517725258493354e-05,
      "loss": 0.3427,
      "step": 2052
    },
    {
      "epoch": 0.75,
      "learning_rate": 3.1508493353028065e-05,
      "loss": 0.2392,
      "step": 2053
    },
    {
      "epoch": 0.75,
      "learning_rate": 3.1499261447562775e-05,
      "loss": 0.4052,
      "step": 2054
    },
    {
      "epoch": 0.75,
      "learning_rate": 3.1490029542097486e-05,
      "loss": 0.0051,
      "step": 2055
    },
    {
      "epoch": 0.75,
      "learning_rate": 3.14807976366322e-05,
      "loss": 0.0568,
      "step": 2056
    },
    {
      "epoch": 0.75,
      "learning_rate": 3.1471565731166914e-05,
      "loss": 0.2934,
      "step": 2057
    },
    {
      "epoch": 0.75,
      "learning_rate": 3.146233382570163e-05,
      "loss": 0.0332,
      "step": 2058
    },
    {
      "epoch": 0.75,
      "learning_rate": 3.145310192023634e-05,
      "loss": 0.1057,
      "step": 2059
    },
    {
      "epoch": 0.75,
      "learning_rate": 3.1443870014771045e-05,
      "loss": 0.1665,
      "step": 2060
    },
    {
      "epoch": 0.75,
      "learning_rate": 3.143463810930576e-05,
      "loss": 0.2789,
      "step": 2061
    },
    {
      "epoch": 0.75,
      "learning_rate": 3.142540620384047e-05,
      "loss": 0.1894,
      "step": 2062
    },
    {
      "epoch": 0.75,
      "learning_rate": 3.1416174298375184e-05,
      "loss": 0.3782,
      "step": 2063
    },
    {
      "epoch": 0.76,
      "learning_rate": 3.14069423929099e-05,
      "loss": 0.2043,
      "step": 2064
    },
    {
      "epoch": 0.76,
      "learning_rate": 3.139771048744461e-05,
      "loss": 0.012,
      "step": 2065
    },
    {
      "epoch": 0.76,
      "learning_rate": 3.138847858197932e-05,
      "loss": 0.021,
      "step": 2066
    },
    {
      "epoch": 0.76,
      "learning_rate": 3.137924667651403e-05,
      "loss": 0.1311,
      "step": 2067
    },
    {
      "epoch": 0.76,
      "learning_rate": 3.137001477104874e-05,
      "loss": 0.0247,
      "step": 2068
    },
    {
      "epoch": 0.76,
      "learning_rate": 3.136078286558346e-05,
      "loss": 0.0255,
      "step": 2069
    },
    {
      "epoch": 0.76,
      "learning_rate": 3.135155096011817e-05,
      "loss": 0.0237,
      "step": 2070
    },
    {
      "epoch": 0.76,
      "learning_rate": 3.134231905465288e-05,
      "loss": 0.4366,
      "step": 2071
    },
    {
      "epoch": 0.76,
      "learning_rate": 3.133308714918759e-05,
      "loss": 0.1484,
      "step": 2072
    },
    {
      "epoch": 0.76,
      "learning_rate": 3.13238552437223e-05,
      "loss": 0.2725,
      "step": 2073
    },
    {
      "epoch": 0.76,
      "learning_rate": 3.131462333825701e-05,
      "loss": 0.4486,
      "step": 2074
    },
    {
      "epoch": 0.76,
      "learning_rate": 3.130539143279173e-05,
      "loss": 0.4802,
      "step": 2075
    },
    {
      "epoch": 0.76,
      "learning_rate": 3.129615952732644e-05,
      "loss": 0.202,
      "step": 2076
    },
    {
      "epoch": 0.76,
      "learning_rate": 3.128692762186116e-05,
      "loss": 0.0776,
      "step": 2077
    },
    {
      "epoch": 0.76,
      "learning_rate": 3.127769571639587e-05,
      "loss": 0.0368,
      "step": 2078
    },
    {
      "epoch": 0.76,
      "learning_rate": 3.126846381093057e-05,
      "loss": 0.0298,
      "step": 2079
    },
    {
      "epoch": 0.76,
      "learning_rate": 3.125923190546529e-05,
      "loss": 0.1378,
      "step": 2080
    },
    {
      "epoch": 0.76,
      "learning_rate": 3.125e-05,
      "loss": 0.4075,
      "step": 2081
    },
    {
      "epoch": 0.76,
      "learning_rate": 3.124076809453471e-05,
      "loss": 0.0032,
      "step": 2082
    },
    {
      "epoch": 0.76,
      "learning_rate": 3.123153618906943e-05,
      "loss": 0.1538,
      "step": 2083
    },
    {
      "epoch": 0.76,
      "learning_rate": 3.122230428360414e-05,
      "loss": 0.0063,
      "step": 2084
    },
    {
      "epoch": 0.76,
      "learning_rate": 3.121307237813885e-05,
      "loss": 0.2414,
      "step": 2085
    },
    {
      "epoch": 0.76,
      "learning_rate": 3.120384047267356e-05,
      "loss": 0.022,
      "step": 2086
    },
    {
      "epoch": 0.76,
      "learning_rate": 3.119460856720827e-05,
      "loss": 0.2952,
      "step": 2087
    },
    {
      "epoch": 0.76,
      "learning_rate": 3.118537666174299e-05,
      "loss": 0.0113,
      "step": 2088
    },
    {
      "epoch": 0.76,
      "learning_rate": 3.11761447562777e-05,
      "loss": 0.017,
      "step": 2089
    },
    {
      "epoch": 0.76,
      "learning_rate": 3.116691285081241e-05,
      "loss": 0.1928,
      "step": 2090
    },
    {
      "epoch": 0.77,
      "learning_rate": 3.115768094534712e-05,
      "loss": 0.2621,
      "step": 2091
    },
    {
      "epoch": 0.77,
      "learning_rate": 3.114844903988183e-05,
      "loss": 0.0389,
      "step": 2092
    },
    {
      "epoch": 0.77,
      "learning_rate": 3.113921713441654e-05,
      "loss": 0.3018,
      "step": 2093
    },
    {
      "epoch": 0.77,
      "learning_rate": 3.112998522895126e-05,
      "loss": 0.0167,
      "step": 2094
    },
    {
      "epoch": 0.77,
      "learning_rate": 3.112075332348597e-05,
      "loss": 0.032,
      "step": 2095
    },
    {
      "epoch": 0.77,
      "learning_rate": 3.111152141802068e-05,
      "loss": 0.0101,
      "step": 2096
    },
    {
      "epoch": 0.77,
      "learning_rate": 3.1102289512555397e-05,
      "loss": 0.0439,
      "step": 2097
    },
    {
      "epoch": 0.77,
      "learning_rate": 3.10930576070901e-05,
      "loss": 0.0422,
      "step": 2098
    },
    {
      "epoch": 0.77,
      "learning_rate": 3.108382570162482e-05,
      "loss": 0.114,
      "step": 2099
    },
    {
      "epoch": 0.77,
      "learning_rate": 3.107459379615953e-05,
      "loss": 0.1397,
      "step": 2100
    },
    {
      "epoch": 0.77,
      "learning_rate": 3.106536189069424e-05,
      "loss": 0.0448,
      "step": 2101
    },
    {
      "epoch": 0.77,
      "learning_rate": 3.1056129985228956e-05,
      "loss": 0.0187,
      "step": 2102
    },
    {
      "epoch": 0.77,
      "learning_rate": 3.1046898079763667e-05,
      "loss": 0.043,
      "step": 2103
    },
    {
      "epoch": 0.77,
      "learning_rate": 3.103766617429838e-05,
      "loss": 0.0773,
      "step": 2104
    },
    {
      "epoch": 0.77,
      "learning_rate": 3.102843426883309e-05,
      "loss": 0.193,
      "step": 2105
    },
    {
      "epoch": 0.77,
      "learning_rate": 3.10192023633678e-05,
      "loss": 0.0103,
      "step": 2106
    },
    {
      "epoch": 0.77,
      "learning_rate": 3.1009970457902516e-05,
      "loss": 0.356,
      "step": 2107
    },
    {
      "epoch": 0.77,
      "learning_rate": 3.1000738552437226e-05,
      "loss": 0.2617,
      "step": 2108
    },
    {
      "epoch": 0.77,
      "learning_rate": 3.099150664697194e-05,
      "loss": 0.4177,
      "step": 2109
    },
    {
      "epoch": 0.77,
      "learning_rate": 3.098227474150665e-05,
      "loss": 0.31,
      "step": 2110
    },
    {
      "epoch": 0.77,
      "learning_rate": 3.097304283604136e-05,
      "loss": 0.1433,
      "step": 2111
    },
    {
      "epoch": 0.77,
      "learning_rate": 3.096381093057607e-05,
      "loss": 0.0061,
      "step": 2112
    },
    {
      "epoch": 0.77,
      "learning_rate": 3.0954579025110786e-05,
      "loss": 0.0145,
      "step": 2113
    },
    {
      "epoch": 0.77,
      "learning_rate": 3.0945347119645496e-05,
      "loss": 0.0415,
      "step": 2114
    },
    {
      "epoch": 0.77,
      "learning_rate": 3.093611521418021e-05,
      "loss": 0.0122,
      "step": 2115
    },
    {
      "epoch": 0.77,
      "learning_rate": 3.0926883308714924e-05,
      "loss": 0.0702,
      "step": 2116
    },
    {
      "epoch": 0.77,
      "learning_rate": 3.091765140324963e-05,
      "loss": 0.0465,
      "step": 2117
    },
    {
      "epoch": 0.77,
      "learning_rate": 3.0908419497784345e-05,
      "loss": 0.2462,
      "step": 2118
    },
    {
      "epoch": 0.78,
      "learning_rate": 3.0899187592319056e-05,
      "loss": 0.0495,
      "step": 2119
    },
    {
      "epoch": 0.78,
      "learning_rate": 3.0889955686853766e-05,
      "loss": 0.3216,
      "step": 2120
    },
    {
      "epoch": 0.78,
      "learning_rate": 3.0880723781388484e-05,
      "loss": 0.0062,
      "step": 2121
    },
    {
      "epoch": 0.78,
      "learning_rate": 3.0871491875923194e-05,
      "loss": 0.0185,
      "step": 2122
    },
    {
      "epoch": 0.78,
      "learning_rate": 3.0862259970457905e-05,
      "loss": 0.0388,
      "step": 2123
    },
    {
      "epoch": 0.78,
      "learning_rate": 3.0853028064992615e-05,
      "loss": 0.0107,
      "step": 2124
    },
    {
      "epoch": 0.78,
      "learning_rate": 3.0843796159527326e-05,
      "loss": 0.2128,
      "step": 2125
    },
    {
      "epoch": 0.78,
      "learning_rate": 3.0834564254062036e-05,
      "loss": 0.0142,
      "step": 2126
    },
    {
      "epoch": 0.78,
      "learning_rate": 3.0825332348596754e-05,
      "loss": 0.365,
      "step": 2127
    },
    {
      "epoch": 0.78,
      "learning_rate": 3.0816100443131464e-05,
      "loss": 0.0127,
      "step": 2128
    },
    {
      "epoch": 0.78,
      "learning_rate": 3.0806868537666175e-05,
      "loss": 0.0286,
      "step": 2129
    },
    {
      "epoch": 0.78,
      "learning_rate": 3.0797636632200885e-05,
      "loss": 0.0064,
      "step": 2130
    },
    {
      "epoch": 0.78,
      "learning_rate": 3.0788404726735596e-05,
      "loss": 0.0054,
      "step": 2131
    },
    {
      "epoch": 0.78,
      "learning_rate": 3.077917282127031e-05,
      "loss": 0.0054,
      "step": 2132
    },
    {
      "epoch": 0.78,
      "learning_rate": 3.0769940915805024e-05,
      "loss": 0.0085,
      "step": 2133
    },
    {
      "epoch": 0.78,
      "learning_rate": 3.0760709010339734e-05,
      "loss": 0.0092,
      "step": 2134
    },
    {
      "epoch": 0.78,
      "learning_rate": 3.075147710487445e-05,
      "loss": 0.0033,
      "step": 2135
    },
    {
      "epoch": 0.78,
      "learning_rate": 3.0742245199409155e-05,
      "loss": 0.0366,
      "step": 2136
    },
    {
      "epoch": 0.78,
      "learning_rate": 3.073301329394387e-05,
      "loss": 0.0032,
      "step": 2137
    },
    {
      "epoch": 0.78,
      "learning_rate": 3.072378138847858e-05,
      "loss": 0.0283,
      "step": 2138
    },
    {
      "epoch": 0.78,
      "learning_rate": 3.0714549483013294e-05,
      "loss": 0.0018,
      "step": 2139
    },
    {
      "epoch": 0.78,
      "learning_rate": 3.070531757754801e-05,
      "loss": 0.0177,
      "step": 2140
    },
    {
      "epoch": 0.78,
      "learning_rate": 3.069608567208272e-05,
      "loss": 0.2442,
      "step": 2141
    },
    {
      "epoch": 0.78,
      "learning_rate": 3.068685376661743e-05,
      "loss": 0.2179,
      "step": 2142
    },
    {
      "epoch": 0.78,
      "learning_rate": 3.067762186115214e-05,
      "loss": 0.2511,
      "step": 2143
    },
    {
      "epoch": 0.78,
      "learning_rate": 3.066838995568685e-05,
      "loss": 0.0231,
      "step": 2144
    },
    {
      "epoch": 0.78,
      "learning_rate": 3.0659158050221564e-05,
      "loss": 0.3251,
      "step": 2145
    },
    {
      "epoch": 0.79,
      "learning_rate": 3.064992614475628e-05,
      "loss": 0.0052,
      "step": 2146
    },
    {
      "epoch": 0.79,
      "learning_rate": 3.064069423929099e-05,
      "loss": 0.1514,
      "step": 2147
    },
    {
      "epoch": 0.79,
      "learning_rate": 3.06314623338257e-05,
      "loss": 0.0084,
      "step": 2148
    },
    {
      "epoch": 0.79,
      "learning_rate": 3.062223042836041e-05,
      "loss": 0.0022,
      "step": 2149
    },
    {
      "epoch": 0.79,
      "learning_rate": 3.0612998522895123e-05,
      "loss": 0.0131,
      "step": 2150
    },
    {
      "epoch": 0.79,
      "learning_rate": 3.060376661742984e-05,
      "loss": 0.0067,
      "step": 2151
    },
    {
      "epoch": 0.79,
      "learning_rate": 3.059453471196455e-05,
      "loss": 0.0066,
      "step": 2152
    },
    {
      "epoch": 0.79,
      "learning_rate": 3.058530280649926e-05,
      "loss": 0.0066,
      "step": 2153
    },
    {
      "epoch": 0.79,
      "learning_rate": 3.057607090103398e-05,
      "loss": 0.4067,
      "step": 2154
    },
    {
      "epoch": 0.79,
      "learning_rate": 3.056683899556868e-05,
      "loss": 0.0048,
      "step": 2155
    },
    {
      "epoch": 0.79,
      "learning_rate": 3.0557607090103393e-05,
      "loss": 0.034,
      "step": 2156
    },
    {
      "epoch": 0.79,
      "learning_rate": 3.054837518463811e-05,
      "loss": 0.241,
      "step": 2157
    },
    {
      "epoch": 0.79,
      "learning_rate": 3.053914327917282e-05,
      "loss": 0.0017,
      "step": 2158
    },
    {
      "epoch": 0.79,
      "learning_rate": 3.052991137370754e-05,
      "loss": 0.0027,
      "step": 2159
    },
    {
      "epoch": 0.79,
      "learning_rate": 3.052067946824225e-05,
      "loss": 0.1517,
      "step": 2160
    },
    {
      "epoch": 0.79,
      "learning_rate": 3.0511447562776956e-05,
      "loss": 0.0031,
      "step": 2161
    },
    {
      "epoch": 0.79,
      "learning_rate": 3.0502215657311674e-05,
      "loss": 0.0012,
      "step": 2162
    },
    {
      "epoch": 0.79,
      "learning_rate": 3.049298375184638e-05,
      "loss": 0.0026,
      "step": 2163
    },
    {
      "epoch": 0.79,
      "learning_rate": 3.048375184638109e-05,
      "loss": 0.3394,
      "step": 2164
    },
    {
      "epoch": 0.79,
      "learning_rate": 3.047451994091581e-05,
      "loss": 0.0075,
      "step": 2165
    },
    {
      "epoch": 0.79,
      "learning_rate": 3.046528803545052e-05,
      "loss": 0.0579,
      "step": 2166
    },
    {
      "epoch": 0.79,
      "learning_rate": 3.0456056129985233e-05,
      "loss": 0.0048,
      "step": 2167
    },
    {
      "epoch": 0.79,
      "learning_rate": 3.0446824224519944e-05,
      "loss": 0.1674,
      "step": 2168
    },
    {
      "epoch": 0.79,
      "learning_rate": 3.0437592319054654e-05,
      "loss": 0.004,
      "step": 2169
    },
    {
      "epoch": 0.79,
      "learning_rate": 3.0428360413589368e-05,
      "loss": 0.1527,
      "step": 2170
    },
    {
      "epoch": 0.79,
      "learning_rate": 3.041912850812408e-05,
      "loss": 0.2045,
      "step": 2171
    },
    {
      "epoch": 0.79,
      "learning_rate": 3.040989660265879e-05,
      "loss": 0.0018,
      "step": 2172
    },
    {
      "epoch": 0.8,
      "learning_rate": 3.0400664697193503e-05,
      "loss": 0.0026,
      "step": 2173
    },
    {
      "epoch": 0.8,
      "learning_rate": 3.0391432791728214e-05,
      "loss": 0.0649,
      "step": 2174
    },
    {
      "epoch": 0.8,
      "learning_rate": 3.0382200886262924e-05,
      "loss": 0.0041,
      "step": 2175
    },
    {
      "epoch": 0.8,
      "learning_rate": 3.037296898079764e-05,
      "loss": 0.0039,
      "step": 2176
    },
    {
      "epoch": 0.8,
      "learning_rate": 3.036373707533235e-05,
      "loss": 0.0103,
      "step": 2177
    },
    {
      "epoch": 0.8,
      "learning_rate": 3.0354505169867066e-05,
      "loss": 0.2843,
      "step": 2178
    },
    {
      "epoch": 0.8,
      "learning_rate": 3.0345273264401773e-05,
      "loss": 0.0075,
      "step": 2179
    },
    {
      "epoch": 0.8,
      "learning_rate": 3.0336041358936484e-05,
      "loss": 0.1166,
      "step": 2180
    },
    {
      "epoch": 0.8,
      "learning_rate": 3.03268094534712e-05,
      "loss": 0.0376,
      "step": 2181
    },
    {
      "epoch": 0.8,
      "learning_rate": 3.031757754800591e-05,
      "loss": 0.0053,
      "step": 2182
    },
    {
      "epoch": 0.8,
      "learning_rate": 3.030834564254062e-05,
      "loss": 0.0032,
      "step": 2183
    },
    {
      "epoch": 0.8,
      "learning_rate": 3.0299113737075336e-05,
      "loss": 0.208,
      "step": 2184
    },
    {
      "epoch": 0.8,
      "learning_rate": 3.0289881831610047e-05,
      "loss": 0.0427,
      "step": 2185
    },
    {
      "epoch": 0.8,
      "learning_rate": 3.0280649926144754e-05,
      "loss": 0.0022,
      "step": 2186
    },
    {
      "epoch": 0.8,
      "learning_rate": 3.027141802067947e-05,
      "loss": 0.0064,
      "step": 2187
    },
    {
      "epoch": 0.8,
      "learning_rate": 3.0262186115214182e-05,
      "loss": 0.0093,
      "step": 2188
    },
    {
      "epoch": 0.8,
      "learning_rate": 3.0252954209748896e-05,
      "loss": 0.0052,
      "step": 2189
    },
    {
      "epoch": 0.8,
      "learning_rate": 3.0243722304283606e-05,
      "loss": 0.0011,
      "step": 2190
    },
    {
      "epoch": 0.8,
      "learning_rate": 3.0234490398818317e-05,
      "loss": 0.0044,
      "step": 2191
    },
    {
      "epoch": 0.8,
      "learning_rate": 3.022525849335303e-05,
      "loss": 0.0015,
      "step": 2192
    },
    {
      "epoch": 0.8,
      "learning_rate": 3.021602658788774e-05,
      "loss": 0.0077,
      "step": 2193
    },
    {
      "epoch": 0.8,
      "learning_rate": 3.0206794682422452e-05,
      "loss": 0.0097,
      "step": 2194
    },
    {
      "epoch": 0.8,
      "learning_rate": 3.0197562776957166e-05,
      "loss": 0.9546,
      "step": 2195
    },
    {
      "epoch": 0.8,
      "learning_rate": 3.0188330871491876e-05,
      "loss": 0.0044,
      "step": 2196
    },
    {
      "epoch": 0.8,
      "learning_rate": 3.0179098966026587e-05,
      "loss": 0.009,
      "step": 2197
    },
    {
      "epoch": 0.8,
      "learning_rate": 3.01698670605613e-05,
      "loss": 0.5925,
      "step": 2198
    },
    {
      "epoch": 0.8,
      "learning_rate": 3.016063515509601e-05,
      "loss": 0.0105,
      "step": 2199
    },
    {
      "epoch": 0.8,
      "learning_rate": 3.015140324963073e-05,
      "loss": 0.0044,
      "step": 2200
    },
    {
      "epoch": 0.81,
      "learning_rate": 3.0142171344165436e-05,
      "loss": 0.0152,
      "step": 2201
    },
    {
      "epoch": 0.81,
      "learning_rate": 3.0132939438700146e-05,
      "loss": 0.3542,
      "step": 2202
    },
    {
      "epoch": 0.81,
      "learning_rate": 3.0123707533234864e-05,
      "loss": 0.0021,
      "step": 2203
    },
    {
      "epoch": 0.81,
      "learning_rate": 3.0114475627769574e-05,
      "loss": 0.2255,
      "step": 2204
    },
    {
      "epoch": 0.81,
      "learning_rate": 3.010524372230428e-05,
      "loss": 0.4787,
      "step": 2205
    },
    {
      "epoch": 0.81,
      "learning_rate": 3.0096011816839e-05,
      "loss": 0.2081,
      "step": 2206
    },
    {
      "epoch": 0.81,
      "learning_rate": 3.008677991137371e-05,
      "loss": 0.463,
      "step": 2207
    },
    {
      "epoch": 0.81,
      "learning_rate": 3.0077548005908423e-05,
      "loss": 0.1402,
      "step": 2208
    },
    {
      "epoch": 0.81,
      "learning_rate": 3.0068316100443134e-05,
      "loss": 0.0016,
      "step": 2209
    },
    {
      "epoch": 0.81,
      "learning_rate": 3.0059084194977844e-05,
      "loss": 0.5751,
      "step": 2210
    },
    {
      "epoch": 0.81,
      "learning_rate": 3.004985228951256e-05,
      "loss": 0.22,
      "step": 2211
    },
    {
      "epoch": 0.81,
      "learning_rate": 3.004062038404727e-05,
      "loss": 0.0062,
      "step": 2212
    },
    {
      "epoch": 0.81,
      "learning_rate": 3.003138847858198e-05,
      "loss": 0.2858,
      "step": 2213
    },
    {
      "epoch": 0.81,
      "learning_rate": 3.0022156573116693e-05,
      "loss": 0.0052,
      "step": 2214
    },
    {
      "epoch": 0.81,
      "learning_rate": 3.0012924667651404e-05,
      "loss": 0.0023,
      "step": 2215
    },
    {
      "epoch": 0.81,
      "learning_rate": 3.0003692762186115e-05,
      "loss": 0.1187,
      "step": 2216
    },
    {
      "epoch": 0.81,
      "learning_rate": 2.999446085672083e-05,
      "loss": 0.293,
      "step": 2217
    },
    {
      "epoch": 0.81,
      "learning_rate": 2.998522895125554e-05,
      "loss": 0.0324,
      "step": 2218
    },
    {
      "epoch": 0.81,
      "learning_rate": 2.9975997045790256e-05,
      "loss": 0.2715,
      "step": 2219
    },
    {
      "epoch": 0.81,
      "learning_rate": 2.9966765140324963e-05,
      "loss": 0.0044,
      "step": 2220
    },
    {
      "epoch": 0.81,
      "learning_rate": 2.9957533234859674e-05,
      "loss": 0.0066,
      "step": 2221
    },
    {
      "epoch": 0.81,
      "learning_rate": 2.994830132939439e-05,
      "loss": 0.1583,
      "step": 2222
    },
    {
      "epoch": 0.81,
      "learning_rate": 2.9939069423929102e-05,
      "loss": 0.0042,
      "step": 2223
    },
    {
      "epoch": 0.81,
      "learning_rate": 2.992983751846381e-05,
      "loss": 0.0043,
      "step": 2224
    },
    {
      "epoch": 0.81,
      "learning_rate": 2.9920605612998526e-05,
      "loss": 0.0035,
      "step": 2225
    },
    {
      "epoch": 0.81,
      "learning_rate": 2.9911373707533237e-05,
      "loss": 0.0036,
      "step": 2226
    },
    {
      "epoch": 0.81,
      "learning_rate": 2.9902141802067944e-05,
      "loss": 0.0051,
      "step": 2227
    },
    {
      "epoch": 0.82,
      "learning_rate": 2.989290989660266e-05,
      "loss": 0.0048,
      "step": 2228
    },
    {
      "epoch": 0.82,
      "learning_rate": 2.9883677991137372e-05,
      "loss": 0.0101,
      "step": 2229
    },
    {
      "epoch": 0.82,
      "learning_rate": 2.9874446085672086e-05,
      "loss": 0.2314,
      "step": 2230
    },
    {
      "epoch": 0.82,
      "learning_rate": 2.9865214180206796e-05,
      "loss": 0.0029,
      "step": 2231
    },
    {
      "epoch": 0.82,
      "learning_rate": 2.9855982274741507e-05,
      "loss": 0.379,
      "step": 2232
    },
    {
      "epoch": 0.82,
      "learning_rate": 2.984675036927622e-05,
      "loss": 0.0038,
      "step": 2233
    },
    {
      "epoch": 0.82,
      "learning_rate": 2.983751846381093e-05,
      "loss": 0.0096,
      "step": 2234
    },
    {
      "epoch": 0.82,
      "learning_rate": 2.9828286558345642e-05,
      "loss": 0.2918,
      "step": 2235
    },
    {
      "epoch": 0.82,
      "learning_rate": 2.9819054652880356e-05,
      "loss": 0.0086,
      "step": 2236
    },
    {
      "epoch": 0.82,
      "learning_rate": 2.9809822747415067e-05,
      "loss": 0.0028,
      "step": 2237
    },
    {
      "epoch": 0.82,
      "learning_rate": 2.9800590841949784e-05,
      "loss": 0.0659,
      "step": 2238
    },
    {
      "epoch": 0.82,
      "learning_rate": 2.979135893648449e-05,
      "loss": 0.0023,
      "step": 2239
    },
    {
      "epoch": 0.82,
      "learning_rate": 2.97821270310192e-05,
      "loss": 0.0059,
      "step": 2240
    },
    {
      "epoch": 0.82,
      "learning_rate": 2.977289512555392e-05,
      "loss": 0.29,
      "step": 2241
    },
    {
      "epoch": 0.82,
      "learning_rate": 2.976366322008863e-05,
      "loss": 0.0172,
      "step": 2242
    },
    {
      "epoch": 0.82,
      "learning_rate": 2.9754431314623337e-05,
      "loss": 0.0636,
      "step": 2243
    },
    {
      "epoch": 0.82,
      "learning_rate": 2.9745199409158054e-05,
      "loss": 0.0605,
      "step": 2244
    },
    {
      "epoch": 0.82,
      "learning_rate": 2.9735967503692764e-05,
      "loss": 0.5488,
      "step": 2245
    },
    {
      "epoch": 0.82,
      "learning_rate": 2.972673559822747e-05,
      "loss": 0.1969,
      "step": 2246
    },
    {
      "epoch": 0.82,
      "learning_rate": 2.971750369276219e-05,
      "loss": 0.0448,
      "step": 2247
    },
    {
      "epoch": 0.82,
      "learning_rate": 2.97082717872969e-05,
      "loss": 0.0046,
      "step": 2248
    },
    {
      "epoch": 0.82,
      "learning_rate": 2.9699039881831613e-05,
      "loss": 0.5595,
      "step": 2249
    },
    {
      "epoch": 0.82,
      "learning_rate": 2.9689807976366324e-05,
      "loss": 0.0184,
      "step": 2250
    },
    {
      "epoch": 0.82,
      "learning_rate": 2.9680576070901035e-05,
      "loss": 0.0229,
      "step": 2251
    },
    {
      "epoch": 0.82,
      "learning_rate": 2.967134416543575e-05,
      "loss": 0.0061,
      "step": 2252
    },
    {
      "epoch": 0.82,
      "learning_rate": 2.966211225997046e-05,
      "loss": 0.2105,
      "step": 2253
    },
    {
      "epoch": 0.82,
      "learning_rate": 2.965288035450517e-05,
      "loss": 0.0084,
      "step": 2254
    },
    {
      "epoch": 0.83,
      "learning_rate": 2.9643648449039884e-05,
      "loss": 0.005,
      "step": 2255
    },
    {
      "epoch": 0.83,
      "learning_rate": 2.9634416543574594e-05,
      "loss": 0.014,
      "step": 2256
    },
    {
      "epoch": 0.83,
      "learning_rate": 2.9625184638109305e-05,
      "loss": 0.0985,
      "step": 2257
    },
    {
      "epoch": 0.83,
      "learning_rate": 2.961595273264402e-05,
      "loss": 0.0282,
      "step": 2258
    },
    {
      "epoch": 0.83,
      "learning_rate": 2.960672082717873e-05,
      "loss": 0.0031,
      "step": 2259
    },
    {
      "epoch": 0.83,
      "learning_rate": 2.9597488921713446e-05,
      "loss": 0.2859,
      "step": 2260
    },
    {
      "epoch": 0.83,
      "learning_rate": 2.9588257016248157e-05,
      "loss": 0.4612,
      "step": 2261
    },
    {
      "epoch": 0.83,
      "learning_rate": 2.9579025110782864e-05,
      "loss": 0.0732,
      "step": 2262
    },
    {
      "epoch": 0.83,
      "learning_rate": 2.956979320531758e-05,
      "loss": 0.2343,
      "step": 2263
    },
    {
      "epoch": 0.83,
      "learning_rate": 2.9560561299852292e-05,
      "loss": 0.0436,
      "step": 2264
    },
    {
      "epoch": 0.83,
      "learning_rate": 2.9551329394387e-05,
      "loss": 0.0061,
      "step": 2265
    },
    {
      "epoch": 0.83,
      "learning_rate": 2.9542097488921716e-05,
      "loss": 0.1395,
      "step": 2266
    },
    {
      "epoch": 0.83,
      "learning_rate": 2.9532865583456427e-05,
      "loss": 0.2273,
      "step": 2267
    },
    {
      "epoch": 0.83,
      "learning_rate": 2.952363367799114e-05,
      "loss": 0.2525,
      "step": 2268
    },
    {
      "epoch": 0.83,
      "learning_rate": 2.951440177252585e-05,
      "loss": 0.0237,
      "step": 2269
    },
    {
      "epoch": 0.83,
      "learning_rate": 2.9505169867060562e-05,
      "loss": 0.0384,
      "step": 2270
    },
    {
      "epoch": 0.83,
      "learning_rate": 2.9495937961595276e-05,
      "loss": 0.3037,
      "step": 2271
    },
    {
      "epoch": 0.83,
      "learning_rate": 2.9486706056129987e-05,
      "loss": 0.0451,
      "step": 2272
    },
    {
      "epoch": 0.83,
      "learning_rate": 2.9477474150664697e-05,
      "loss": 0.0258,
      "step": 2273
    },
    {
      "epoch": 0.83,
      "learning_rate": 2.946824224519941e-05,
      "loss": 0.1707,
      "step": 2274
    },
    {
      "epoch": 0.83,
      "learning_rate": 2.945901033973412e-05,
      "loss": 0.0149,
      "step": 2275
    },
    {
      "epoch": 0.83,
      "learning_rate": 2.9449778434268832e-05,
      "loss": 0.0154,
      "step": 2276
    },
    {
      "epoch": 0.83,
      "learning_rate": 2.9440546528803546e-05,
      "loss": 0.2785,
      "step": 2277
    },
    {
      "epoch": 0.83,
      "learning_rate": 2.9431314623338257e-05,
      "loss": 0.0043,
      "step": 2278
    },
    {
      "epoch": 0.83,
      "learning_rate": 2.9422082717872974e-05,
      "loss": 0.0246,
      "step": 2279
    },
    {
      "epoch": 0.83,
      "learning_rate": 2.9412850812407685e-05,
      "loss": 0.0732,
      "step": 2280
    },
    {
      "epoch": 0.83,
      "learning_rate": 2.940361890694239e-05,
      "loss": 0.0237,
      "step": 2281
    },
    {
      "epoch": 0.83,
      "learning_rate": 2.939438700147711e-05,
      "loss": 0.0112,
      "step": 2282
    },
    {
      "epoch": 0.84,
      "learning_rate": 2.938515509601182e-05,
      "loss": 0.0296,
      "step": 2283
    },
    {
      "epoch": 0.84,
      "learning_rate": 2.9375923190546527e-05,
      "loss": 0.1092,
      "step": 2284
    },
    {
      "epoch": 0.84,
      "learning_rate": 2.9366691285081244e-05,
      "loss": 0.2185,
      "step": 2285
    },
    {
      "epoch": 0.84,
      "learning_rate": 2.9357459379615955e-05,
      "loss": 0.0087,
      "step": 2286
    },
    {
      "epoch": 0.84,
      "learning_rate": 2.9348227474150665e-05,
      "loss": 0.3403,
      "step": 2287
    },
    {
      "epoch": 0.84,
      "learning_rate": 2.933899556868538e-05,
      "loss": 0.5297,
      "step": 2288
    },
    {
      "epoch": 0.84,
      "learning_rate": 2.932976366322009e-05,
      "loss": 0.0039,
      "step": 2289
    },
    {
      "epoch": 0.84,
      "learning_rate": 2.9320531757754804e-05,
      "loss": 0.005,
      "step": 2290
    },
    {
      "epoch": 0.84,
      "learning_rate": 2.9311299852289514e-05,
      "loss": 0.0084,
      "step": 2291
    },
    {
      "epoch": 0.84,
      "learning_rate": 2.9302067946824225e-05,
      "loss": 0.1342,
      "step": 2292
    },
    {
      "epoch": 0.84,
      "learning_rate": 2.929283604135894e-05,
      "loss": 0.4284,
      "step": 2293
    },
    {
      "epoch": 0.84,
      "learning_rate": 2.928360413589365e-05,
      "loss": 0.3383,
      "step": 2294
    },
    {
      "epoch": 0.84,
      "learning_rate": 2.927437223042836e-05,
      "loss": 0.4345,
      "step": 2295
    },
    {
      "epoch": 0.84,
      "learning_rate": 2.9265140324963074e-05,
      "loss": 0.0263,
      "step": 2296
    },
    {
      "epoch": 0.84,
      "learning_rate": 2.9255908419497784e-05,
      "loss": 0.0046,
      "step": 2297
    },
    {
      "epoch": 0.84,
      "learning_rate": 2.9246676514032495e-05,
      "loss": 0.014,
      "step": 2298
    },
    {
      "epoch": 0.84,
      "learning_rate": 2.9237444608567212e-05,
      "loss": 0.0074,
      "step": 2299
    },
    {
      "epoch": 0.84,
      "learning_rate": 2.922821270310192e-05,
      "loss": 0.0365,
      "step": 2300
    },
    {
      "epoch": 0.84,
      "learning_rate": 2.9218980797636637e-05,
      "loss": 0.0431,
      "step": 2301
    },
    {
      "epoch": 0.84,
      "learning_rate": 2.9209748892171347e-05,
      "loss": 0.0077,
      "step": 2302
    },
    {
      "epoch": 0.84,
      "learning_rate": 2.9200516986706054e-05,
      "loss": 0.013,
      "step": 2303
    },
    {
      "epoch": 0.84,
      "learning_rate": 2.919128508124077e-05,
      "loss": 0.0089,
      "step": 2304
    },
    {
      "epoch": 0.84,
      "learning_rate": 2.9182053175775482e-05,
      "loss": 0.1299,
      "step": 2305
    },
    {
      "epoch": 0.84,
      "learning_rate": 2.9172821270310193e-05,
      "loss": 0.0139,
      "step": 2306
    },
    {
      "epoch": 0.84,
      "learning_rate": 2.9163589364844907e-05,
      "loss": 0.014,
      "step": 2307
    },
    {
      "epoch": 0.84,
      "learning_rate": 2.9154357459379617e-05,
      "loss": 0.0395,
      "step": 2308
    },
    {
      "epoch": 0.84,
      "learning_rate": 2.914512555391433e-05,
      "loss": 0.0308,
      "step": 2309
    },
    {
      "epoch": 0.85,
      "learning_rate": 2.913589364844904e-05,
      "loss": 0.0045,
      "step": 2310
    },
    {
      "epoch": 0.85,
      "learning_rate": 2.9126661742983752e-05,
      "loss": 0.4366,
      "step": 2311
    },
    {
      "epoch": 0.85,
      "learning_rate": 2.9117429837518466e-05,
      "loss": 0.0033,
      "step": 2312
    },
    {
      "epoch": 0.85,
      "learning_rate": 2.9108197932053177e-05,
      "loss": 0.2618,
      "step": 2313
    },
    {
      "epoch": 0.85,
      "learning_rate": 2.9098966026587887e-05,
      "loss": 0.1816,
      "step": 2314
    },
    {
      "epoch": 0.85,
      "learning_rate": 2.90897341211226e-05,
      "loss": 0.001,
      "step": 2315
    },
    {
      "epoch": 0.85,
      "learning_rate": 2.9080502215657312e-05,
      "loss": 0.2632,
      "step": 2316
    },
    {
      "epoch": 0.85,
      "learning_rate": 2.9071270310192022e-05,
      "loss": 0.0027,
      "step": 2317
    },
    {
      "epoch": 0.85,
      "learning_rate": 2.906203840472674e-05,
      "loss": 0.1264,
      "step": 2318
    },
    {
      "epoch": 0.85,
      "learning_rate": 2.9052806499261447e-05,
      "loss": 0.0495,
      "step": 2319
    },
    {
      "epoch": 0.85,
      "learning_rate": 2.9043574593796164e-05,
      "loss": 0.0063,
      "step": 2320
    },
    {
      "epoch": 0.85,
      "learning_rate": 2.9034342688330875e-05,
      "loss": 0.2532,
      "step": 2321
    },
    {
      "epoch": 0.85,
      "learning_rate": 2.9025110782865582e-05,
      "loss": 0.0025,
      "step": 2322
    },
    {
      "epoch": 0.85,
      "learning_rate": 2.90158788774003e-05,
      "loss": 0.0304,
      "step": 2323
    },
    {
      "epoch": 0.85,
      "learning_rate": 2.900664697193501e-05,
      "loss": 0.0018,
      "step": 2324
    },
    {
      "epoch": 0.85,
      "learning_rate": 2.899741506646972e-05,
      "loss": 0.4876,
      "step": 2325
    },
    {
      "epoch": 0.85,
      "learning_rate": 2.8988183161004434e-05,
      "loss": 0.5024,
      "step": 2326
    },
    {
      "epoch": 0.85,
      "learning_rate": 2.8978951255539145e-05,
      "loss": 0.3516,
      "step": 2327
    },
    {
      "epoch": 0.85,
      "learning_rate": 2.8969719350073855e-05,
      "loss": 0.001,
      "step": 2328
    },
    {
      "epoch": 0.85,
      "learning_rate": 2.896048744460857e-05,
      "loss": 0.2849,
      "step": 2329
    },
    {
      "epoch": 0.85,
      "learning_rate": 2.895125553914328e-05,
      "loss": 0.1784,
      "step": 2330
    },
    {
      "epoch": 0.85,
      "learning_rate": 2.8942023633677994e-05,
      "loss": 0.0064,
      "step": 2331
    },
    {
      "epoch": 0.85,
      "learning_rate": 2.8932791728212704e-05,
      "loss": 0.0027,
      "step": 2332
    },
    {
      "epoch": 0.85,
      "learning_rate": 2.8923559822747415e-05,
      "loss": 0.0238,
      "step": 2333
    },
    {
      "epoch": 0.85,
      "learning_rate": 2.891432791728213e-05,
      "loss": 0.0017,
      "step": 2334
    },
    {
      "epoch": 0.85,
      "learning_rate": 2.890509601181684e-05,
      "loss": 0.088,
      "step": 2335
    },
    {
      "epoch": 0.85,
      "learning_rate": 2.889586410635155e-05,
      "loss": 0.1613,
      "step": 2336
    },
    {
      "epoch": 0.86,
      "learning_rate": 2.8886632200886267e-05,
      "loss": 0.0034,
      "step": 2337
    },
    {
      "epoch": 0.86,
      "learning_rate": 2.8877400295420974e-05,
      "loss": 0.006,
      "step": 2338
    },
    {
      "epoch": 0.86,
      "learning_rate": 2.886816838995569e-05,
      "loss": 0.0018,
      "step": 2339
    },
    {
      "epoch": 0.86,
      "learning_rate": 2.8858936484490402e-05,
      "loss": 0.0607,
      "step": 2340
    },
    {
      "epoch": 0.86,
      "learning_rate": 2.884970457902511e-05,
      "loss": 0.0046,
      "step": 2341
    },
    {
      "epoch": 0.86,
      "learning_rate": 2.8840472673559827e-05,
      "loss": 0.1683,
      "step": 2342
    },
    {
      "epoch": 0.86,
      "learning_rate": 2.8831240768094537e-05,
      "loss": 0.0068,
      "step": 2343
    },
    {
      "epoch": 0.86,
      "learning_rate": 2.8822008862629248e-05,
      "loss": 0.0293,
      "step": 2344
    },
    {
      "epoch": 0.86,
      "learning_rate": 2.881277695716396e-05,
      "loss": 0.0061,
      "step": 2345
    },
    {
      "epoch": 0.86,
      "learning_rate": 2.8803545051698672e-05,
      "loss": 0.0124,
      "step": 2346
    },
    {
      "epoch": 0.86,
      "learning_rate": 2.8794313146233383e-05,
      "loss": 0.0126,
      "step": 2347
    },
    {
      "epoch": 0.86,
      "learning_rate": 2.8785081240768097e-05,
      "loss": 0.0018,
      "step": 2348
    },
    {
      "epoch": 0.86,
      "learning_rate": 2.8775849335302807e-05,
      "loss": 0.3316,
      "step": 2349
    },
    {
      "epoch": 0.86,
      "learning_rate": 2.876661742983752e-05,
      "loss": 0.0518,
      "step": 2350
    },
    {
      "epoch": 0.86,
      "learning_rate": 2.8757385524372232e-05,
      "loss": 0.0349,
      "step": 2351
    },
    {
      "epoch": 0.86,
      "learning_rate": 2.8748153618906942e-05,
      "loss": 0.0025,
      "step": 2352
    },
    {
      "epoch": 0.86,
      "learning_rate": 2.8738921713441656e-05,
      "loss": 0.2058,
      "step": 2353
    },
    {
      "epoch": 0.86,
      "learning_rate": 2.8729689807976367e-05,
      "loss": 0.0022,
      "step": 2354
    },
    {
      "epoch": 0.86,
      "learning_rate": 2.8720457902511077e-05,
      "loss": 0.0612,
      "step": 2355
    },
    {
      "epoch": 0.86,
      "learning_rate": 2.8711225997045795e-05,
      "loss": 0.0019,
      "step": 2356
    },
    {
      "epoch": 0.86,
      "learning_rate": 2.8701994091580502e-05,
      "loss": 0.006,
      "step": 2357
    },
    {
      "epoch": 0.86,
      "learning_rate": 2.8692762186115212e-05,
      "loss": 0.0022,
      "step": 2358
    },
    {
      "epoch": 0.86,
      "learning_rate": 2.868353028064993e-05,
      "loss": 0.1825,
      "step": 2359
    },
    {
      "epoch": 0.86,
      "learning_rate": 2.8674298375184637e-05,
      "loss": 0.0066,
      "step": 2360
    },
    {
      "epoch": 0.86,
      "learning_rate": 2.8665066469719354e-05,
      "loss": 0.0022,
      "step": 2361
    },
    {
      "epoch": 0.86,
      "learning_rate": 2.8655834564254065e-05,
      "loss": 0.2158,
      "step": 2362
    },
    {
      "epoch": 0.86,
      "learning_rate": 2.8646602658788775e-05,
      "loss": 0.0017,
      "step": 2363
    },
    {
      "epoch": 0.86,
      "learning_rate": 2.863737075332349e-05,
      "loss": 0.0015,
      "step": 2364
    },
    {
      "epoch": 0.87,
      "learning_rate": 2.86281388478582e-05,
      "loss": 0.0034,
      "step": 2365
    },
    {
      "epoch": 0.87,
      "learning_rate": 2.861890694239291e-05,
      "loss": 0.0089,
      "step": 2366
    },
    {
      "epoch": 0.87,
      "learning_rate": 2.8609675036927624e-05,
      "loss": 0.0011,
      "step": 2367
    },
    {
      "epoch": 0.87,
      "learning_rate": 2.8600443131462335e-05,
      "loss": 0.1414,
      "step": 2368
    },
    {
      "epoch": 0.87,
      "learning_rate": 2.859121122599705e-05,
      "loss": 0.0317,
      "step": 2369
    },
    {
      "epoch": 0.87,
      "learning_rate": 2.858197932053176e-05,
      "loss": 0.3487,
      "step": 2370
    },
    {
      "epoch": 0.87,
      "learning_rate": 2.857274741506647e-05,
      "loss": 0.0011,
      "step": 2371
    },
    {
      "epoch": 0.87,
      "learning_rate": 2.8563515509601184e-05,
      "loss": 0.0262,
      "step": 2372
    },
    {
      "epoch": 0.87,
      "learning_rate": 2.8554283604135894e-05,
      "loss": 0.0906,
      "step": 2373
    },
    {
      "epoch": 0.87,
      "learning_rate": 2.8545051698670605e-05,
      "loss": 0.0121,
      "step": 2374
    },
    {
      "epoch": 0.87,
      "learning_rate": 2.8535819793205322e-05,
      "loss": 0.0062,
      "step": 2375
    },
    {
      "epoch": 0.87,
      "learning_rate": 2.852658788774003e-05,
      "loss": 0.0045,
      "step": 2376
    },
    {
      "epoch": 0.87,
      "learning_rate": 2.851735598227474e-05,
      "loss": 0.2366,
      "step": 2377
    },
    {
      "epoch": 0.87,
      "learning_rate": 2.8508124076809457e-05,
      "loss": 0.2324,
      "step": 2378
    },
    {
      "epoch": 0.87,
      "learning_rate": 2.8498892171344164e-05,
      "loss": 0.0011,
      "step": 2379
    },
    {
      "epoch": 0.87,
      "learning_rate": 2.8489660265878882e-05,
      "loss": 0.1017,
      "step": 2380
    },
    {
      "epoch": 0.87,
      "learning_rate": 2.8480428360413592e-05,
      "loss": 0.0008,
      "step": 2381
    },
    {
      "epoch": 0.87,
      "learning_rate": 2.8471196454948303e-05,
      "loss": 0.0008,
      "step": 2382
    },
    {
      "epoch": 0.87,
      "learning_rate": 2.8461964549483017e-05,
      "loss": 0.0028,
      "step": 2383
    },
    {
      "epoch": 0.87,
      "learning_rate": 2.8452732644017727e-05,
      "loss": 0.0869,
      "step": 2384
    },
    {
      "epoch": 0.87,
      "learning_rate": 2.8443500738552438e-05,
      "loss": 0.0024,
      "step": 2385
    },
    {
      "epoch": 0.87,
      "learning_rate": 2.8434268833087152e-05,
      "loss": 0.0034,
      "step": 2386
    },
    {
      "epoch": 0.87,
      "learning_rate": 2.8425036927621862e-05,
      "loss": 0.0059,
      "step": 2387
    },
    {
      "epoch": 0.87,
      "learning_rate": 2.8415805022156573e-05,
      "loss": 0.0024,
      "step": 2388
    },
    {
      "epoch": 0.87,
      "learning_rate": 2.8406573116691287e-05,
      "loss": 0.0042,
      "step": 2389
    },
    {
      "epoch": 0.87,
      "learning_rate": 2.8397341211225997e-05,
      "loss": 0.254,
      "step": 2390
    },
    {
      "epoch": 0.87,
      "learning_rate": 2.838810930576071e-05,
      "loss": 0.0018,
      "step": 2391
    },
    {
      "epoch": 0.88,
      "learning_rate": 2.8378877400295422e-05,
      "loss": 0.004,
      "step": 2392
    },
    {
      "epoch": 0.88,
      "learning_rate": 2.8369645494830132e-05,
      "loss": 0.4725,
      "step": 2393
    },
    {
      "epoch": 0.88,
      "learning_rate": 2.836041358936485e-05,
      "loss": 0.0087,
      "step": 2394
    },
    {
      "epoch": 0.88,
      "learning_rate": 2.8351181683899557e-05,
      "loss": 0.0038,
      "step": 2395
    },
    {
      "epoch": 0.88,
      "learning_rate": 2.8341949778434267e-05,
      "loss": 0.3798,
      "step": 2396
    },
    {
      "epoch": 0.88,
      "learning_rate": 2.8332717872968985e-05,
      "loss": 0.7656,
      "step": 2397
    },
    {
      "epoch": 0.88,
      "learning_rate": 2.8323485967503692e-05,
      "loss": 0.0009,
      "step": 2398
    },
    {
      "epoch": 0.88,
      "learning_rate": 2.8314254062038402e-05,
      "loss": 0.3231,
      "step": 2399
    },
    {
      "epoch": 0.88,
      "learning_rate": 2.830502215657312e-05,
      "loss": 0.0019,
      "step": 2400
    },
    {
      "epoch": 0.88,
      "learning_rate": 2.829579025110783e-05,
      "loss": 0.0022,
      "step": 2401
    },
    {
      "epoch": 0.88,
      "learning_rate": 2.8286558345642544e-05,
      "loss": 0.3467,
      "step": 2402
    },
    {
      "epoch": 0.88,
      "learning_rate": 2.8277326440177255e-05,
      "loss": 0.0567,
      "step": 2403
    },
    {
      "epoch": 0.88,
      "learning_rate": 2.8268094534711965e-05,
      "loss": 0.2957,
      "step": 2404
    },
    {
      "epoch": 0.88,
      "learning_rate": 2.825886262924668e-05,
      "loss": 0.0035,
      "step": 2405
    },
    {
      "epoch": 0.88,
      "learning_rate": 2.824963072378139e-05,
      "loss": 0.0021,
      "step": 2406
    },
    {
      "epoch": 0.88,
      "learning_rate": 2.82403988183161e-05,
      "loss": 0.2296,
      "step": 2407
    },
    {
      "epoch": 0.88,
      "learning_rate": 2.8231166912850814e-05,
      "loss": 0.6014,
      "step": 2408
    },
    {
      "epoch": 0.88,
      "learning_rate": 2.8221935007385525e-05,
      "loss": 0.0598,
      "step": 2409
    },
    {
      "epoch": 0.88,
      "learning_rate": 2.821270310192024e-05,
      "loss": 0.0034,
      "step": 2410
    },
    {
      "epoch": 0.88,
      "learning_rate": 2.820347119645495e-05,
      "loss": 0.0147,
      "step": 2411
    },
    {
      "epoch": 0.88,
      "learning_rate": 2.819423929098966e-05,
      "loss": 0.0164,
      "step": 2412
    },
    {
      "epoch": 0.88,
      "learning_rate": 2.8185007385524377e-05,
      "loss": 0.0177,
      "step": 2413
    },
    {
      "epoch": 0.88,
      "learning_rate": 2.8175775480059084e-05,
      "loss": 0.004,
      "step": 2414
    },
    {
      "epoch": 0.88,
      "learning_rate": 2.8166543574593795e-05,
      "loss": 0.0065,
      "step": 2415
    },
    {
      "epoch": 0.88,
      "learning_rate": 2.8157311669128512e-05,
      "loss": 0.0304,
      "step": 2416
    },
    {
      "epoch": 0.88,
      "learning_rate": 2.814807976366322e-05,
      "loss": 0.0207,
      "step": 2417
    },
    {
      "epoch": 0.88,
      "learning_rate": 2.813884785819793e-05,
      "loss": 0.0025,
      "step": 2418
    },
    {
      "epoch": 0.89,
      "learning_rate": 2.8129615952732647e-05,
      "loss": 0.2206,
      "step": 2419
    },
    {
      "epoch": 0.89,
      "learning_rate": 2.8120384047267358e-05,
      "loss": 0.002,
      "step": 2420
    },
    {
      "epoch": 0.89,
      "learning_rate": 2.8111152141802072e-05,
      "loss": 0.0035,
      "step": 2421
    },
    {
      "epoch": 0.89,
      "learning_rate": 2.8101920236336782e-05,
      "loss": 0.3715,
      "step": 2422
    },
    {
      "epoch": 0.89,
      "learning_rate": 2.8092688330871493e-05,
      "loss": 0.0017,
      "step": 2423
    },
    {
      "epoch": 0.89,
      "learning_rate": 2.8083456425406207e-05,
      "loss": 0.002,
      "step": 2424
    },
    {
      "epoch": 0.89,
      "learning_rate": 2.8074224519940917e-05,
      "loss": 0.0021,
      "step": 2425
    },
    {
      "epoch": 0.89,
      "learning_rate": 2.8064992614475628e-05,
      "loss": 0.0025,
      "step": 2426
    },
    {
      "epoch": 0.89,
      "learning_rate": 2.8055760709010342e-05,
      "loss": 0.4709,
      "step": 2427
    },
    {
      "epoch": 0.89,
      "learning_rate": 2.8046528803545052e-05,
      "loss": 0.0015,
      "step": 2428
    },
    {
      "epoch": 0.89,
      "learning_rate": 2.8037296898079763e-05,
      "loss": 0.2303,
      "step": 2429
    },
    {
      "epoch": 0.89,
      "learning_rate": 2.8028064992614477e-05,
      "loss": 0.0103,
      "step": 2430
    },
    {
      "epoch": 0.89,
      "learning_rate": 2.8018833087149187e-05,
      "loss": 0.0012,
      "step": 2431
    },
    {
      "epoch": 0.89,
      "learning_rate": 2.8009601181683905e-05,
      "loss": 0.7376,
      "step": 2432
    },
    {
      "epoch": 0.89,
      "learning_rate": 2.8000369276218612e-05,
      "loss": 0.3336,
      "step": 2433
    },
    {
      "epoch": 0.89,
      "learning_rate": 2.7991137370753323e-05,
      "loss": 0.0099,
      "step": 2434
    },
    {
      "epoch": 0.89,
      "learning_rate": 2.798190546528804e-05,
      "loss": 0.1013,
      "step": 2435
    },
    {
      "epoch": 0.89,
      "learning_rate": 2.7972673559822747e-05,
      "loss": 0.2052,
      "step": 2436
    },
    {
      "epoch": 0.89,
      "learning_rate": 2.7963441654357458e-05,
      "loss": 0.0035,
      "step": 2437
    },
    {
      "epoch": 0.89,
      "learning_rate": 2.7954209748892175e-05,
      "loss": 0.0039,
      "step": 2438
    },
    {
      "epoch": 0.89,
      "learning_rate": 2.7944977843426885e-05,
      "loss": 0.0111,
      "step": 2439
    },
    {
      "epoch": 0.89,
      "learning_rate": 2.79357459379616e-05,
      "loss": 0.3034,
      "step": 2440
    },
    {
      "epoch": 0.89,
      "learning_rate": 2.792651403249631e-05,
      "loss": 0.0327,
      "step": 2441
    },
    {
      "epoch": 0.89,
      "learning_rate": 2.791728212703102e-05,
      "loss": 0.0594,
      "step": 2442
    },
    {
      "epoch": 0.89,
      "learning_rate": 2.7908050221565734e-05,
      "loss": 0.3007,
      "step": 2443
    },
    {
      "epoch": 0.89,
      "learning_rate": 2.7898818316100445e-05,
      "loss": 0.0029,
      "step": 2444
    },
    {
      "epoch": 0.89,
      "learning_rate": 2.7889586410635156e-05,
      "loss": 0.0098,
      "step": 2445
    },
    {
      "epoch": 0.89,
      "learning_rate": 2.788035450516987e-05,
      "loss": 0.0006,
      "step": 2446
    },
    {
      "epoch": 0.9,
      "learning_rate": 2.787112259970458e-05,
      "loss": 0.0033,
      "step": 2447
    },
    {
      "epoch": 0.9,
      "learning_rate": 2.786189069423929e-05,
      "loss": 0.0141,
      "step": 2448
    },
    {
      "epoch": 0.9,
      "learning_rate": 2.7852658788774004e-05,
      "loss": 0.0038,
      "step": 2449
    },
    {
      "epoch": 0.9,
      "learning_rate": 2.7843426883308715e-05,
      "loss": 0.0075,
      "step": 2450
    },
    {
      "epoch": 0.9,
      "learning_rate": 2.7834194977843432e-05,
      "loss": 0.0036,
      "step": 2451
    },
    {
      "epoch": 0.9,
      "learning_rate": 2.782496307237814e-05,
      "loss": 0.2416,
      "step": 2452
    },
    {
      "epoch": 0.9,
      "learning_rate": 2.781573116691285e-05,
      "loss": 0.014,
      "step": 2453
    },
    {
      "epoch": 0.9,
      "learning_rate": 2.7806499261447567e-05,
      "loss": 0.0019,
      "step": 2454
    },
    {
      "epoch": 0.9,
      "learning_rate": 2.7797267355982275e-05,
      "loss": 0.0054,
      "step": 2455
    },
    {
      "epoch": 0.9,
      "learning_rate": 2.7788035450516985e-05,
      "loss": 0.0101,
      "step": 2456
    },
    {
      "epoch": 0.9,
      "learning_rate": 2.7778803545051702e-05,
      "loss": 0.7882,
      "step": 2457
    },
    {
      "epoch": 0.9,
      "learning_rate": 2.7769571639586413e-05,
      "loss": 0.0269,
      "step": 2458
    },
    {
      "epoch": 0.9,
      "learning_rate": 2.776033973412112e-05,
      "loss": 0.4243,
      "step": 2459
    },
    {
      "epoch": 0.9,
      "learning_rate": 2.7751107828655837e-05,
      "loss": 0.0012,
      "step": 2460
    },
    {
      "epoch": 0.9,
      "learning_rate": 2.7741875923190548e-05,
      "loss": 0.0075,
      "step": 2461
    },
    {
      "epoch": 0.9,
      "learning_rate": 2.7732644017725262e-05,
      "loss": 0.226,
      "step": 2462
    },
    {
      "epoch": 0.9,
      "learning_rate": 2.7723412112259972e-05,
      "loss": 0.227,
      "step": 2463
    },
    {
      "epoch": 0.9,
      "learning_rate": 2.7714180206794683e-05,
      "loss": 0.0045,
      "step": 2464
    },
    {
      "epoch": 0.9,
      "learning_rate": 2.7704948301329397e-05,
      "loss": 0.0031,
      "step": 2465
    },
    {
      "epoch": 0.9,
      "learning_rate": 2.7695716395864108e-05,
      "loss": 0.0076,
      "step": 2466
    },
    {
      "epoch": 0.9,
      "learning_rate": 2.7686484490398818e-05,
      "loss": 0.2391,
      "step": 2467
    },
    {
      "epoch": 0.9,
      "learning_rate": 2.7677252584933532e-05,
      "loss": 0.5842,
      "step": 2468
    },
    {
      "epoch": 0.9,
      "learning_rate": 2.7668020679468243e-05,
      "loss": 0.0066,
      "step": 2469
    },
    {
      "epoch": 0.9,
      "learning_rate": 2.7658788774002953e-05,
      "loss": 0.1432,
      "step": 2470
    },
    {
      "epoch": 0.9,
      "learning_rate": 2.7649556868537667e-05,
      "loss": 0.0075,
      "step": 2471
    },
    {
      "epoch": 0.9,
      "learning_rate": 2.7640324963072378e-05,
      "loss": 0.0193,
      "step": 2472
    },
    {
      "epoch": 0.9,
      "learning_rate": 2.7631093057607095e-05,
      "loss": 0.0055,
      "step": 2473
    },
    {
      "epoch": 0.91,
      "learning_rate": 2.7621861152141802e-05,
      "loss": 0.0043,
      "step": 2474
    },
    {
      "epoch": 0.91,
      "learning_rate": 2.7612629246676513e-05,
      "loss": 0.0012,
      "step": 2475
    },
    {
      "epoch": 0.91,
      "learning_rate": 2.760339734121123e-05,
      "loss": 0.201,
      "step": 2476
    },
    {
      "epoch": 0.91,
      "learning_rate": 2.759416543574594e-05,
      "loss": 0.0063,
      "step": 2477
    },
    {
      "epoch": 0.91,
      "learning_rate": 2.7584933530280648e-05,
      "loss": 0.0183,
      "step": 2478
    },
    {
      "epoch": 0.91,
      "learning_rate": 2.7575701624815365e-05,
      "loss": 0.0191,
      "step": 2479
    },
    {
      "epoch": 0.91,
      "learning_rate": 2.7566469719350076e-05,
      "loss": 0.2416,
      "step": 2480
    },
    {
      "epoch": 0.91,
      "learning_rate": 2.755723781388479e-05,
      "loss": 0.2477,
      "step": 2481
    },
    {
      "epoch": 0.91,
      "learning_rate": 2.75480059084195e-05,
      "loss": 0.0025,
      "step": 2482
    },
    {
      "epoch": 0.91,
      "learning_rate": 2.753877400295421e-05,
      "loss": 0.0017,
      "step": 2483
    },
    {
      "epoch": 0.91,
      "learning_rate": 2.7529542097488925e-05,
      "loss": 0.004,
      "step": 2484
    },
    {
      "epoch": 0.91,
      "learning_rate": 2.7520310192023635e-05,
      "loss": 0.004,
      "step": 2485
    },
    {
      "epoch": 0.91,
      "learning_rate": 2.7511078286558346e-05,
      "loss": 0.1424,
      "step": 2486
    },
    {
      "epoch": 0.91,
      "learning_rate": 2.750184638109306e-05,
      "loss": 0.2755,
      "step": 2487
    },
    {
      "epoch": 0.91,
      "learning_rate": 2.749261447562777e-05,
      "loss": 0.0033,
      "step": 2488
    },
    {
      "epoch": 0.91,
      "learning_rate": 2.748338257016248e-05,
      "loss": 0.0043,
      "step": 2489
    },
    {
      "epoch": 0.91,
      "learning_rate": 2.7474150664697195e-05,
      "loss": 0.0016,
      "step": 2490
    },
    {
      "epoch": 0.91,
      "learning_rate": 2.7464918759231905e-05,
      "loss": 0.1794,
      "step": 2491
    },
    {
      "epoch": 0.91,
      "learning_rate": 2.7455686853766622e-05,
      "loss": 0.2854,
      "step": 2492
    },
    {
      "epoch": 0.91,
      "learning_rate": 2.744645494830133e-05,
      "loss": 0.053,
      "step": 2493
    },
    {
      "epoch": 0.91,
      "learning_rate": 2.743722304283604e-05,
      "loss": 0.004,
      "step": 2494
    },
    {
      "epoch": 0.91,
      "learning_rate": 2.7427991137370757e-05,
      "loss": 0.6093,
      "step": 2495
    },
    {
      "epoch": 0.91,
      "learning_rate": 2.7418759231905468e-05,
      "loss": 0.0073,
      "step": 2496
    },
    {
      "epoch": 0.91,
      "learning_rate": 2.7409527326440175e-05,
      "loss": 0.3805,
      "step": 2497
    },
    {
      "epoch": 0.91,
      "learning_rate": 2.7400295420974893e-05,
      "loss": 0.0053,
      "step": 2498
    },
    {
      "epoch": 0.91,
      "learning_rate": 2.7391063515509603e-05,
      "loss": 0.009,
      "step": 2499
    },
    {
      "epoch": 0.91,
      "learning_rate": 2.738183161004431e-05,
      "loss": 0.3484,
      "step": 2500
    },
    {
      "epoch": 0.92,
      "learning_rate": 2.7372599704579028e-05,
      "loss": 0.4085,
      "step": 2501
    },
    {
      "epoch": 0.92,
      "learning_rate": 2.7363367799113738e-05,
      "loss": 0.0078,
      "step": 2502
    },
    {
      "epoch": 0.92,
      "learning_rate": 2.7354135893648452e-05,
      "loss": 0.0798,
      "step": 2503
    },
    {
      "epoch": 0.92,
      "learning_rate": 2.7344903988183163e-05,
      "loss": 0.0659,
      "step": 2504
    },
    {
      "epoch": 0.92,
      "learning_rate": 2.7335672082717873e-05,
      "loss": 0.1687,
      "step": 2505
    },
    {
      "epoch": 0.92,
      "learning_rate": 2.7326440177252587e-05,
      "loss": 0.0065,
      "step": 2506
    },
    {
      "epoch": 0.92,
      "learning_rate": 2.7317208271787298e-05,
      "loss": 0.0051,
      "step": 2507
    },
    {
      "epoch": 0.92,
      "learning_rate": 2.7307976366322008e-05,
      "loss": 0.0031,
      "step": 2508
    },
    {
      "epoch": 0.92,
      "learning_rate": 2.7298744460856722e-05,
      "loss": 0.0076,
      "step": 2509
    },
    {
      "epoch": 0.92,
      "learning_rate": 2.7289512555391433e-05,
      "loss": 0.5733,
      "step": 2510
    },
    {
      "epoch": 0.92,
      "learning_rate": 2.728028064992615e-05,
      "loss": 0.0072,
      "step": 2511
    },
    {
      "epoch": 0.92,
      "learning_rate": 2.7271048744460857e-05,
      "loss": 0.1542,
      "step": 2512
    },
    {
      "epoch": 0.92,
      "learning_rate": 2.7261816838995568e-05,
      "loss": 0.0423,
      "step": 2513
    },
    {
      "epoch": 0.92,
      "learning_rate": 2.7252584933530285e-05,
      "loss": 0.0083,
      "step": 2514
    },
    {
      "epoch": 0.92,
      "learning_rate": 2.7243353028064996e-05,
      "loss": 0.0881,
      "step": 2515
    },
    {
      "epoch": 0.92,
      "learning_rate": 2.7234121122599703e-05,
      "loss": 0.0273,
      "step": 2516
    },
    {
      "epoch": 0.92,
      "learning_rate": 2.722488921713442e-05,
      "loss": 0.3072,
      "step": 2517
    },
    {
      "epoch": 0.92,
      "learning_rate": 2.721565731166913e-05,
      "loss": 0.0079,
      "step": 2518
    },
    {
      "epoch": 0.92,
      "learning_rate": 2.7206425406203838e-05,
      "loss": 0.0044,
      "step": 2519
    },
    {
      "epoch": 0.92,
      "learning_rate": 2.7197193500738555e-05,
      "loss": 0.004,
      "step": 2520
    },
    {
      "epoch": 0.92,
      "learning_rate": 2.7187961595273266e-05,
      "loss": 0.4328,
      "step": 2521
    },
    {
      "epoch": 0.92,
      "learning_rate": 2.717872968980798e-05,
      "loss": 0.0053,
      "step": 2522
    },
    {
      "epoch": 0.92,
      "learning_rate": 2.716949778434269e-05,
      "loss": 0.0119,
      "step": 2523
    },
    {
      "epoch": 0.92,
      "learning_rate": 2.71602658788774e-05,
      "loss": 0.0102,
      "step": 2524
    },
    {
      "epoch": 0.92,
      "learning_rate": 2.7151033973412115e-05,
      "loss": 0.3056,
      "step": 2525
    },
    {
      "epoch": 0.92,
      "learning_rate": 2.7141802067946825e-05,
      "loss": 0.0072,
      "step": 2526
    },
    {
      "epoch": 0.92,
      "learning_rate": 2.7132570162481536e-05,
      "loss": 0.3063,
      "step": 2527
    },
    {
      "epoch": 0.92,
      "learning_rate": 2.712333825701625e-05,
      "loss": 0.0061,
      "step": 2528
    },
    {
      "epoch": 0.93,
      "learning_rate": 2.711410635155096e-05,
      "loss": 0.3392,
      "step": 2529
    },
    {
      "epoch": 0.93,
      "learning_rate": 2.710487444608567e-05,
      "loss": 0.0226,
      "step": 2530
    },
    {
      "epoch": 0.93,
      "learning_rate": 2.7095642540620385e-05,
      "loss": 0.0112,
      "step": 2531
    },
    {
      "epoch": 0.93,
      "learning_rate": 2.7086410635155095e-05,
      "loss": 0.0017,
      "step": 2532
    },
    {
      "epoch": 0.93,
      "learning_rate": 2.7077178729689813e-05,
      "loss": 0.0058,
      "step": 2533
    },
    {
      "epoch": 0.93,
      "learning_rate": 2.7067946824224523e-05,
      "loss": 0.0031,
      "step": 2534
    },
    {
      "epoch": 0.93,
      "learning_rate": 2.705871491875923e-05,
      "loss": 0.2153,
      "step": 2535
    },
    {
      "epoch": 0.93,
      "learning_rate": 2.7049483013293948e-05,
      "loss": 0.0071,
      "step": 2536
    },
    {
      "epoch": 0.93,
      "learning_rate": 2.7040251107828658e-05,
      "loss": 0.0039,
      "step": 2537
    },
    {
      "epoch": 0.93,
      "learning_rate": 2.7031019202363365e-05,
      "loss": 0.4322,
      "step": 2538
    },
    {
      "epoch": 0.93,
      "learning_rate": 2.7021787296898083e-05,
      "loss": 0.0088,
      "step": 2539
    },
    {
      "epoch": 0.93,
      "learning_rate": 2.7012555391432793e-05,
      "loss": 0.0043,
      "step": 2540
    },
    {
      "epoch": 0.93,
      "learning_rate": 2.7003323485967507e-05,
      "loss": 0.0037,
      "step": 2541
    },
    {
      "epoch": 0.93,
      "learning_rate": 2.6994091580502218e-05,
      "loss": 0.4607,
      "step": 2542
    },
    {
      "epoch": 0.93,
      "learning_rate": 2.6984859675036928e-05,
      "loss": 0.0029,
      "step": 2543
    },
    {
      "epoch": 0.93,
      "learning_rate": 2.6975627769571642e-05,
      "loss": 0.0059,
      "step": 2544
    },
    {
      "epoch": 0.93,
      "learning_rate": 2.6966395864106353e-05,
      "loss": 0.1423,
      "step": 2545
    },
    {
      "epoch": 0.93,
      "learning_rate": 2.6957163958641063e-05,
      "loss": 0.0045,
      "step": 2546
    },
    {
      "epoch": 0.93,
      "learning_rate": 2.6947932053175777e-05,
      "loss": 0.0108,
      "step": 2547
    },
    {
      "epoch": 0.93,
      "learning_rate": 2.6938700147710488e-05,
      "loss": 0.0165,
      "step": 2548
    },
    {
      "epoch": 0.93,
      "learning_rate": 2.6929468242245198e-05,
      "loss": 0.0052,
      "step": 2549
    },
    {
      "epoch": 0.93,
      "learning_rate": 2.6920236336779912e-05,
      "loss": 0.0056,
      "step": 2550
    },
    {
      "epoch": 0.93,
      "learning_rate": 2.6911004431314623e-05,
      "loss": 0.0056,
      "step": 2551
    },
    {
      "epoch": 0.93,
      "learning_rate": 2.690177252584934e-05,
      "loss": 0.0035,
      "step": 2552
    },
    {
      "epoch": 0.93,
      "learning_rate": 2.689254062038405e-05,
      "loss": 0.0042,
      "step": 2553
    },
    {
      "epoch": 0.93,
      "learning_rate": 2.6883308714918758e-05,
      "loss": 0.0016,
      "step": 2554
    },
    {
      "epoch": 0.93,
      "learning_rate": 2.6874076809453475e-05,
      "loss": 0.2538,
      "step": 2555
    },
    {
      "epoch": 0.94,
      "learning_rate": 2.6864844903988186e-05,
      "loss": 0.0033,
      "step": 2556
    },
    {
      "epoch": 0.94,
      "learning_rate": 2.6855612998522893e-05,
      "loss": 0.0012,
      "step": 2557
    },
    {
      "epoch": 0.94,
      "learning_rate": 2.684638109305761e-05,
      "loss": 0.365,
      "step": 2558
    },
    {
      "epoch": 0.94,
      "learning_rate": 2.683714918759232e-05,
      "loss": 0.3205,
      "step": 2559
    },
    {
      "epoch": 0.94,
      "learning_rate": 2.682791728212703e-05,
      "loss": 0.0054,
      "step": 2560
    },
    {
      "epoch": 0.94,
      "learning_rate": 2.6818685376661745e-05,
      "loss": 0.0101,
      "step": 2561
    },
    {
      "epoch": 0.94,
      "learning_rate": 2.6809453471196456e-05,
      "loss": 0.0026,
      "step": 2562
    },
    {
      "epoch": 0.94,
      "learning_rate": 2.680022156573117e-05,
      "loss": 0.0032,
      "step": 2563
    },
    {
      "epoch": 0.94,
      "learning_rate": 2.679098966026588e-05,
      "loss": 0.0084,
      "step": 2564
    },
    {
      "epoch": 0.94,
      "learning_rate": 2.678175775480059e-05,
      "loss": 0.0025,
      "step": 2565
    },
    {
      "epoch": 0.94,
      "learning_rate": 2.6772525849335305e-05,
      "loss": 0.0046,
      "step": 2566
    },
    {
      "epoch": 0.94,
      "learning_rate": 2.6763293943870015e-05,
      "loss": 0.005,
      "step": 2567
    },
    {
      "epoch": 0.94,
      "learning_rate": 2.6754062038404726e-05,
      "loss": 0.0041,
      "step": 2568
    },
    {
      "epoch": 0.94,
      "learning_rate": 2.674483013293944e-05,
      "loss": 0.253,
      "step": 2569
    },
    {
      "epoch": 0.94,
      "learning_rate": 2.673559822747415e-05,
      "loss": 0.6003,
      "step": 2570
    },
    {
      "epoch": 0.94,
      "learning_rate": 2.672636632200886e-05,
      "loss": 0.0031,
      "step": 2571
    },
    {
      "epoch": 0.94,
      "learning_rate": 2.6717134416543578e-05,
      "loss": 0.0068,
      "step": 2572
    },
    {
      "epoch": 0.94,
      "learning_rate": 2.6707902511078285e-05,
      "loss": 0.2527,
      "step": 2573
    },
    {
      "epoch": 0.94,
      "learning_rate": 2.6698670605613003e-05,
      "loss": 0.0644,
      "step": 2574
    },
    {
      "epoch": 0.94,
      "learning_rate": 2.6689438700147713e-05,
      "loss": 0.1159,
      "step": 2575
    },
    {
      "epoch": 0.94,
      "learning_rate": 2.668020679468242e-05,
      "loss": 0.2157,
      "step": 2576
    },
    {
      "epoch": 0.94,
      "learning_rate": 2.6670974889217138e-05,
      "loss": 0.2207,
      "step": 2577
    },
    {
      "epoch": 0.94,
      "learning_rate": 2.6661742983751848e-05,
      "loss": 0.0058,
      "step": 2578
    },
    {
      "epoch": 0.94,
      "learning_rate": 2.665251107828656e-05,
      "loss": 0.0118,
      "step": 2579
    },
    {
      "epoch": 0.94,
      "learning_rate": 2.6643279172821273e-05,
      "loss": 0.2749,
      "step": 2580
    },
    {
      "epoch": 0.94,
      "learning_rate": 2.6634047267355983e-05,
      "loss": 0.2148,
      "step": 2581
    },
    {
      "epoch": 0.94,
      "learning_rate": 2.6624815361890697e-05,
      "loss": 0.0086,
      "step": 2582
    },
    {
      "epoch": 0.95,
      "learning_rate": 2.6615583456425408e-05,
      "loss": 0.0111,
      "step": 2583
    },
    {
      "epoch": 0.95,
      "learning_rate": 2.660635155096012e-05,
      "loss": 0.0035,
      "step": 2584
    },
    {
      "epoch": 0.95,
      "learning_rate": 2.6597119645494832e-05,
      "loss": 0.0091,
      "step": 2585
    },
    {
      "epoch": 0.95,
      "learning_rate": 2.6587887740029543e-05,
      "loss": 0.004,
      "step": 2586
    },
    {
      "epoch": 0.95,
      "learning_rate": 2.6578655834564253e-05,
      "loss": 0.021,
      "step": 2587
    },
    {
      "epoch": 0.95,
      "learning_rate": 2.6569423929098967e-05,
      "loss": 0.336,
      "step": 2588
    },
    {
      "epoch": 0.95,
      "learning_rate": 2.6560192023633678e-05,
      "loss": 0.0084,
      "step": 2589
    },
    {
      "epoch": 0.95,
      "learning_rate": 2.655096011816839e-05,
      "loss": 0.1772,
      "step": 2590
    },
    {
      "epoch": 0.95,
      "learning_rate": 2.6541728212703106e-05,
      "loss": 0.7144,
      "step": 2591
    },
    {
      "epoch": 0.95,
      "learning_rate": 2.6532496307237813e-05,
      "loss": 0.051,
      "step": 2592
    },
    {
      "epoch": 0.95,
      "learning_rate": 2.652326440177253e-05,
      "loss": 0.0169,
      "step": 2593
    },
    {
      "epoch": 0.95,
      "learning_rate": 2.651403249630724e-05,
      "loss": 0.3748,
      "step": 2594
    },
    {
      "epoch": 0.95,
      "learning_rate": 2.6504800590841948e-05,
      "loss": 0.0029,
      "step": 2595
    },
    {
      "epoch": 0.95,
      "learning_rate": 2.6495568685376665e-05,
      "loss": 0.342,
      "step": 2596
    },
    {
      "epoch": 0.95,
      "learning_rate": 2.6486336779911376e-05,
      "loss": 0.4405,
      "step": 2597
    },
    {
      "epoch": 0.95,
      "learning_rate": 2.6477104874446086e-05,
      "loss": 0.1739,
      "step": 2598
    },
    {
      "epoch": 0.95,
      "learning_rate": 2.64678729689808e-05,
      "loss": 0.2462,
      "step": 2599
    },
    {
      "epoch": 0.95,
      "learning_rate": 2.645864106351551e-05,
      "loss": 0.0072,
      "step": 2600
    },
    {
      "epoch": 0.95,
      "learning_rate": 2.644940915805022e-05,
      "loss": 0.0031,
      "step": 2601
    },
    {
      "epoch": 0.95,
      "learning_rate": 2.6440177252584935e-05,
      "loss": 0.01,
      "step": 2602
    },
    {
      "epoch": 0.95,
      "learning_rate": 2.6430945347119646e-05,
      "loss": 0.0048,
      "step": 2603
    },
    {
      "epoch": 0.95,
      "learning_rate": 2.642171344165436e-05,
      "loss": 0.0069,
      "step": 2604
    },
    {
      "epoch": 0.95,
      "learning_rate": 2.641248153618907e-05,
      "loss": 0.0108,
      "step": 2605
    },
    {
      "epoch": 0.95,
      "learning_rate": 2.640324963072378e-05,
      "loss": 0.0062,
      "step": 2606
    },
    {
      "epoch": 0.95,
      "learning_rate": 2.6394017725258495e-05,
      "loss": 0.0162,
      "step": 2607
    },
    {
      "epoch": 0.95,
      "learning_rate": 2.6384785819793205e-05,
      "loss": 0.0034,
      "step": 2608
    },
    {
      "epoch": 0.95,
      "learning_rate": 2.6375553914327916e-05,
      "loss": 0.0091,
      "step": 2609
    },
    {
      "epoch": 0.95,
      "learning_rate": 2.6366322008862633e-05,
      "loss": 0.0051,
      "step": 2610
    },
    {
      "epoch": 0.96,
      "learning_rate": 2.635709010339734e-05,
      "loss": 0.0164,
      "step": 2611
    },
    {
      "epoch": 0.96,
      "learning_rate": 2.6347858197932058e-05,
      "loss": 0.0023,
      "step": 2612
    },
    {
      "epoch": 0.96,
      "learning_rate": 2.6338626292466768e-05,
      "loss": 0.0025,
      "step": 2613
    },
    {
      "epoch": 0.96,
      "learning_rate": 2.6329394387001475e-05,
      "loss": 0.5115,
      "step": 2614
    },
    {
      "epoch": 0.96,
      "learning_rate": 2.6320162481536193e-05,
      "loss": 0.0039,
      "step": 2615
    },
    {
      "epoch": 0.96,
      "learning_rate": 2.6310930576070903e-05,
      "loss": 0.3561,
      "step": 2616
    },
    {
      "epoch": 0.96,
      "learning_rate": 2.6301698670605614e-05,
      "loss": 0.0033,
      "step": 2617
    },
    {
      "epoch": 0.96,
      "learning_rate": 2.6292466765140328e-05,
      "loss": 0.0021,
      "step": 2618
    },
    {
      "epoch": 0.96,
      "learning_rate": 2.628323485967504e-05,
      "loss": 0.0024,
      "step": 2619
    },
    {
      "epoch": 0.96,
      "learning_rate": 2.627400295420975e-05,
      "loss": 0.0034,
      "step": 2620
    },
    {
      "epoch": 0.96,
      "learning_rate": 2.6264771048744463e-05,
      "loss": 0.2616,
      "step": 2621
    },
    {
      "epoch": 0.96,
      "learning_rate": 2.6255539143279173e-05,
      "loss": 0.007,
      "step": 2622
    },
    {
      "epoch": 0.96,
      "learning_rate": 2.6246307237813887e-05,
      "loss": 0.0023,
      "step": 2623
    },
    {
      "epoch": 0.96,
      "learning_rate": 2.6237075332348598e-05,
      "loss": 0.3789,
      "step": 2624
    },
    {
      "epoch": 0.96,
      "learning_rate": 2.622784342688331e-05,
      "loss": 0.0132,
      "step": 2625
    },
    {
      "epoch": 0.96,
      "learning_rate": 2.6218611521418022e-05,
      "loss": 0.0057,
      "step": 2626
    },
    {
      "epoch": 0.96,
      "learning_rate": 2.6209379615952733e-05,
      "loss": 0.3088,
      "step": 2627
    },
    {
      "epoch": 0.96,
      "learning_rate": 2.6200147710487443e-05,
      "loss": 0.1422,
      "step": 2628
    },
    {
      "epoch": 0.96,
      "learning_rate": 2.619091580502216e-05,
      "loss": 0.129,
      "step": 2629
    },
    {
      "epoch": 0.96,
      "learning_rate": 2.6181683899556868e-05,
      "loss": 0.0096,
      "step": 2630
    },
    {
      "epoch": 0.96,
      "learning_rate": 2.617245199409158e-05,
      "loss": 0.237,
      "step": 2631
    },
    {
      "epoch": 0.96,
      "learning_rate": 2.6163220088626296e-05,
      "loss": 0.0032,
      "step": 2632
    },
    {
      "epoch": 0.96,
      "learning_rate": 2.6153988183161003e-05,
      "loss": 0.0451,
      "step": 2633
    },
    {
      "epoch": 0.96,
      "learning_rate": 2.614475627769572e-05,
      "loss": 0.0098,
      "step": 2634
    },
    {
      "epoch": 0.96,
      "learning_rate": 2.613552437223043e-05,
      "loss": 0.0079,
      "step": 2635
    },
    {
      "epoch": 0.96,
      "learning_rate": 2.612629246676514e-05,
      "loss": 0.045,
      "step": 2636
    },
    {
      "epoch": 0.96,
      "learning_rate": 2.6117060561299855e-05,
      "loss": 0.4561,
      "step": 2637
    },
    {
      "epoch": 0.97,
      "learning_rate": 2.6107828655834566e-05,
      "loss": 0.0066,
      "step": 2638
    },
    {
      "epoch": 0.97,
      "learning_rate": 2.6098596750369276e-05,
      "loss": 0.0105,
      "step": 2639
    },
    {
      "epoch": 0.97,
      "learning_rate": 2.608936484490399e-05,
      "loss": 0.3271,
      "step": 2640
    },
    {
      "epoch": 0.97,
      "learning_rate": 2.60801329394387e-05,
      "loss": 0.0035,
      "step": 2641
    },
    {
      "epoch": 0.97,
      "learning_rate": 2.6070901033973415e-05,
      "loss": 0.007,
      "step": 2642
    },
    {
      "epoch": 0.97,
      "learning_rate": 2.6061669128508125e-05,
      "loss": 0.0026,
      "step": 2643
    },
    {
      "epoch": 0.97,
      "learning_rate": 2.6052437223042836e-05,
      "loss": 0.0917,
      "step": 2644
    },
    {
      "epoch": 0.97,
      "learning_rate": 2.604320531757755e-05,
      "loss": 0.0017,
      "step": 2645
    },
    {
      "epoch": 0.97,
      "learning_rate": 2.603397341211226e-05,
      "loss": 0.0043,
      "step": 2646
    },
    {
      "epoch": 0.97,
      "learning_rate": 2.602474150664697e-05,
      "loss": 0.2953,
      "step": 2647
    },
    {
      "epoch": 0.97,
      "learning_rate": 2.601550960118169e-05,
      "loss": 0.0079,
      "step": 2648
    },
    {
      "epoch": 0.97,
      "learning_rate": 2.6006277695716395e-05,
      "loss": 0.0335,
      "step": 2649
    },
    {
      "epoch": 0.97,
      "learning_rate": 2.5997045790251106e-05,
      "loss": 0.0039,
      "step": 2650
    },
    {
      "epoch": 0.97,
      "learning_rate": 2.5987813884785823e-05,
      "loss": 0.0025,
      "step": 2651
    },
    {
      "epoch": 0.97,
      "learning_rate": 2.597858197932053e-05,
      "loss": 0.1374,
      "step": 2652
    },
    {
      "epoch": 0.97,
      "learning_rate": 2.5969350073855248e-05,
      "loss": 0.0032,
      "step": 2653
    },
    {
      "epoch": 0.97,
      "learning_rate": 2.596011816838996e-05,
      "loss": 0.0096,
      "step": 2654
    },
    {
      "epoch": 0.97,
      "learning_rate": 2.595088626292467e-05,
      "loss": 0.0023,
      "step": 2655
    },
    {
      "epoch": 0.97,
      "learning_rate": 2.5941654357459383e-05,
      "loss": 0.2851,
      "step": 2656
    },
    {
      "epoch": 0.97,
      "learning_rate": 2.5932422451994093e-05,
      "loss": 0.0019,
      "step": 2657
    },
    {
      "epoch": 0.97,
      "learning_rate": 2.5923190546528804e-05,
      "loss": 0.0963,
      "step": 2658
    },
    {
      "epoch": 0.97,
      "learning_rate": 2.5913958641063518e-05,
      "loss": 0.3088,
      "step": 2659
    },
    {
      "epoch": 0.97,
      "learning_rate": 2.590472673559823e-05,
      "loss": 0.0023,
      "step": 2660
    },
    {
      "epoch": 0.97,
      "learning_rate": 2.589549483013294e-05,
      "loss": 0.0048,
      "step": 2661
    },
    {
      "epoch": 0.97,
      "learning_rate": 2.5886262924667653e-05,
      "loss": 0.176,
      "step": 2662
    },
    {
      "epoch": 0.97,
      "learning_rate": 2.5877031019202364e-05,
      "loss": 0.1825,
      "step": 2663
    },
    {
      "epoch": 0.97,
      "learning_rate": 2.5867799113737077e-05,
      "loss": 0.0065,
      "step": 2664
    },
    {
      "epoch": 0.98,
      "learning_rate": 2.5858567208271788e-05,
      "loss": 0.0029,
      "step": 2665
    },
    {
      "epoch": 0.98,
      "learning_rate": 2.58493353028065e-05,
      "loss": 0.0017,
      "step": 2666
    },
    {
      "epoch": 0.98,
      "learning_rate": 2.5840103397341216e-05,
      "loss": 0.3154,
      "step": 2667
    },
    {
      "epoch": 0.98,
      "learning_rate": 2.5830871491875923e-05,
      "loss": 0.2653,
      "step": 2668
    },
    {
      "epoch": 0.98,
      "learning_rate": 2.5821639586410634e-05,
      "loss": 0.1445,
      "step": 2669
    },
    {
      "epoch": 0.98,
      "learning_rate": 2.581240768094535e-05,
      "loss": 0.1459,
      "step": 2670
    },
    {
      "epoch": 0.98,
      "learning_rate": 2.5803175775480058e-05,
      "loss": 0.0023,
      "step": 2671
    },
    {
      "epoch": 0.98,
      "learning_rate": 2.579394387001477e-05,
      "loss": 0.0157,
      "step": 2672
    },
    {
      "epoch": 0.98,
      "learning_rate": 2.5784711964549486e-05,
      "loss": 0.0382,
      "step": 2673
    },
    {
      "epoch": 0.98,
      "learning_rate": 2.5775480059084196e-05,
      "loss": 0.0066,
      "step": 2674
    },
    {
      "epoch": 0.98,
      "learning_rate": 2.576624815361891e-05,
      "loss": 0.2495,
      "step": 2675
    },
    {
      "epoch": 0.98,
      "learning_rate": 2.575701624815362e-05,
      "loss": 0.177,
      "step": 2676
    },
    {
      "epoch": 0.98,
      "learning_rate": 2.574778434268833e-05,
      "loss": 0.1301,
      "step": 2677
    },
    {
      "epoch": 0.98,
      "learning_rate": 2.5738552437223045e-05,
      "loss": 0.0347,
      "step": 2678
    },
    {
      "epoch": 0.98,
      "learning_rate": 2.5729320531757756e-05,
      "loss": 0.0052,
      "step": 2679
    },
    {
      "epoch": 0.98,
      "learning_rate": 2.5720088626292467e-05,
      "loss": 0.0219,
      "step": 2680
    },
    {
      "epoch": 0.98,
      "learning_rate": 2.571085672082718e-05,
      "loss": 0.0141,
      "step": 2681
    },
    {
      "epoch": 0.98,
      "learning_rate": 2.570162481536189e-05,
      "loss": 0.3284,
      "step": 2682
    },
    {
      "epoch": 0.98,
      "learning_rate": 2.5692392909896605e-05,
      "loss": 0.015,
      "step": 2683
    },
    {
      "epoch": 0.98,
      "learning_rate": 2.5683161004431316e-05,
      "loss": 0.015,
      "step": 2684
    },
    {
      "epoch": 0.98,
      "learning_rate": 2.5673929098966026e-05,
      "loss": 0.0155,
      "step": 2685
    },
    {
      "epoch": 0.98,
      "learning_rate": 2.5664697193500743e-05,
      "loss": 0.0065,
      "step": 2686
    },
    {
      "epoch": 0.98,
      "learning_rate": 2.565546528803545e-05,
      "loss": 0.0958,
      "step": 2687
    },
    {
      "epoch": 0.98,
      "learning_rate": 2.564623338257016e-05,
      "loss": 0.0065,
      "step": 2688
    },
    {
      "epoch": 0.98,
      "learning_rate": 2.563700147710488e-05,
      "loss": 0.1111,
      "step": 2689
    },
    {
      "epoch": 0.98,
      "learning_rate": 2.5627769571639586e-05,
      "loss": 0.0139,
      "step": 2690
    },
    {
      "epoch": 0.98,
      "learning_rate": 2.5618537666174296e-05,
      "loss": 0.1527,
      "step": 2691
    },
    {
      "epoch": 0.98,
      "learning_rate": 2.5609305760709013e-05,
      "loss": 0.1709,
      "step": 2692
    },
    {
      "epoch": 0.99,
      "learning_rate": 2.5600073855243724e-05,
      "loss": 0.2612,
      "step": 2693
    },
    {
      "epoch": 0.99,
      "learning_rate": 2.5590841949778438e-05,
      "loss": 0.1759,
      "step": 2694
    },
    {
      "epoch": 0.99,
      "learning_rate": 2.558161004431315e-05,
      "loss": 0.0036,
      "step": 2695
    },
    {
      "epoch": 0.99,
      "learning_rate": 2.557237813884786e-05,
      "loss": 0.0039,
      "step": 2696
    },
    {
      "epoch": 0.99,
      "learning_rate": 2.5563146233382573e-05,
      "loss": 0.0134,
      "step": 2697
    },
    {
      "epoch": 0.99,
      "learning_rate": 2.5553914327917284e-05,
      "loss": 0.2777,
      "step": 2698
    },
    {
      "epoch": 0.99,
      "learning_rate": 2.5544682422451994e-05,
      "loss": 0.0119,
      "step": 2699
    },
    {
      "epoch": 0.99,
      "learning_rate": 2.5535450516986708e-05,
      "loss": 0.3059,
      "step": 2700
    },
    {
      "epoch": 0.99,
      "learning_rate": 2.552621861152142e-05,
      "loss": 0.0028,
      "step": 2701
    },
    {
      "epoch": 0.99,
      "learning_rate": 2.551698670605613e-05,
      "loss": 0.0259,
      "step": 2702
    },
    {
      "epoch": 0.99,
      "learning_rate": 2.5507754800590843e-05,
      "loss": 0.2571,
      "step": 2703
    },
    {
      "epoch": 0.99,
      "learning_rate": 2.5498522895125554e-05,
      "loss": 0.0236,
      "step": 2704
    },
    {
      "epoch": 0.99,
      "learning_rate": 2.548929098966027e-05,
      "loss": 0.012,
      "step": 2705
    },
    {
      "epoch": 0.99,
      "learning_rate": 2.5480059084194978e-05,
      "loss": 0.0032,
      "step": 2706
    },
    {
      "epoch": 0.99,
      "learning_rate": 2.547082717872969e-05,
      "loss": 0.2636,
      "step": 2707
    },
    {
      "epoch": 0.99,
      "learning_rate": 2.5461595273264406e-05,
      "loss": 0.2405,
      "step": 2708
    },
    {
      "epoch": 0.99,
      "learning_rate": 2.5452363367799113e-05,
      "loss": 0.1267,
      "step": 2709
    },
    {
      "epoch": 0.99,
      "learning_rate": 2.5443131462333824e-05,
      "loss": 0.0197,
      "step": 2710
    },
    {
      "epoch": 0.99,
      "learning_rate": 2.543389955686854e-05,
      "loss": 0.3337,
      "step": 2711
    },
    {
      "epoch": 0.99,
      "learning_rate": 2.542466765140325e-05,
      "loss": 0.0066,
      "step": 2712
    },
    {
      "epoch": 0.99,
      "learning_rate": 2.5415435745937966e-05,
      "loss": 0.1764,
      "step": 2713
    },
    {
      "epoch": 0.99,
      "learning_rate": 2.5406203840472676e-05,
      "loss": 0.2957,
      "step": 2714
    },
    {
      "epoch": 0.99,
      "learning_rate": 2.5396971935007387e-05,
      "loss": 0.0061,
      "step": 2715
    },
    {
      "epoch": 0.99,
      "learning_rate": 2.53877400295421e-05,
      "loss": 0.0025,
      "step": 2716
    },
    {
      "epoch": 0.99,
      "learning_rate": 2.537850812407681e-05,
      "loss": 0.0035,
      "step": 2717
    },
    {
      "epoch": 0.99,
      "learning_rate": 2.536927621861152e-05,
      "loss": 0.0515,
      "step": 2718
    },
    {
      "epoch": 0.99,
      "learning_rate": 2.5360044313146236e-05,
      "loss": 0.0067,
      "step": 2719
    },
    {
      "epoch": 1.0,
      "learning_rate": 2.5350812407680946e-05,
      "loss": 0.0109,
      "step": 2720
    },
    {
      "epoch": 1.0,
      "learning_rate": 2.5341580502215657e-05,
      "loss": 0.0017,
      "step": 2721
    },
    {
      "epoch": 1.0,
      "learning_rate": 2.533234859675037e-05,
      "loss": 0.2775,
      "step": 2722
    },
    {
      "epoch": 1.0,
      "learning_rate": 2.532311669128508e-05,
      "loss": 0.0085,
      "step": 2723
    },
    {
      "epoch": 1.0,
      "learning_rate": 2.53138847858198e-05,
      "loss": 0.1079,
      "step": 2724
    },
    {
      "epoch": 1.0,
      "learning_rate": 2.5304652880354506e-05,
      "loss": 0.003,
      "step": 2725
    },
    {
      "epoch": 1.0,
      "learning_rate": 2.5295420974889216e-05,
      "loss": 0.0031,
      "step": 2726
    },
    {
      "epoch": 1.0,
      "learning_rate": 2.5286189069423934e-05,
      "loss": 0.1041,
      "step": 2727
    },
    {
      "epoch": 1.0,
      "learning_rate": 2.527695716395864e-05,
      "loss": 0.0072,
      "step": 2728
    },
    {
      "epoch": 1.0,
      "learning_rate": 2.526772525849335e-05,
      "loss": 0.0052,
      "step": 2729
    },
    {
      "epoch": 1.0,
      "learning_rate": 2.525849335302807e-05,
      "loss": 0.1289,
      "step": 2730
    },
    {
      "epoch": 1.0,
      "learning_rate": 2.524926144756278e-05,
      "loss": 0.2946,
      "step": 2731
    },
    {
      "epoch": 1.0,
      "learning_rate": 2.5240029542097486e-05,
      "loss": 0.0041,
      "step": 2732
    },
    {
      "epoch": 1.0,
      "learning_rate": 2.5230797636632204e-05,
      "loss": 0.0023,
      "step": 2733
    },
    {
      "epoch": 1.0,
      "eval_accuracy": 0.9728283244133388,
      "eval_loss": 0.08992622047662735,
      "eval_runtime": 10.853,
      "eval_samples_per_second": 447.616,
      "eval_steps_per_second": 28.011,
      "step": 2733
    }
  ],
  "max_steps": 5466,
  "num_train_epochs": 2,
  "total_flos": 327935464191552.0,
  "trial_name": null,
  "trial_params": null
}