diff --git "a/checkpoint-14000/trainer_state.json" "b/checkpoint-14000/trainer_state.json"
new file mode 100644--- /dev/null
+++ "b/checkpoint-14000/trainer_state.json"
@@ -0,0 +1,84016 @@
+{
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 0.5590561269057408,
+  "global_step": 14000,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.0000000000000001e-07,
+      "loss": 1.0488,
+      "step": 1
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 2.0000000000000002e-07,
+      "loss": 1.1027,
+      "step": 2
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 3.0000000000000004e-07,
+      "loss": 1.2865,
+      "step": 3
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 4.0000000000000003e-07,
+      "loss": 1.0681,
+      "step": 4
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 5.000000000000001e-07,
+      "loss": 1.1824,
+      "step": 5
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 6.000000000000001e-07,
+      "loss": 1.1611,
+      "step": 6
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 7.000000000000001e-07,
+      "loss": 1.0395,
+      "step": 7
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 8.000000000000001e-07,
+      "loss": 1.1625,
+      "step": 8
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 9.000000000000001e-07,
+      "loss": 1.1697,
+      "step": 9
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.0000000000000002e-06,
+      "loss": 1.233,
+      "step": 10
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.1e-06,
+      "loss": 1.1196,
+      "step": 11
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.2000000000000002e-06,
+      "loss": 1.028,
+      "step": 12
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.3e-06,
+      "loss": 1.079,
+      "step": 13
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.4000000000000001e-06,
+      "loss": 1.161,
+      "step": 14
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.5e-06,
+      "loss": 1.1258,
+      "step": 15
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.6000000000000001e-06,
+      "loss": 1.1793,
+      "step": 16
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.7000000000000002e-06,
+      "loss": 1.2195,
+      "step": 17
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.8000000000000001e-06,
+      "loss": 1.0552,
+      "step": 18
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.9000000000000002e-06,
+      "loss": 0.9736,
+      "step": 19
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 2.0000000000000003e-06,
+      "loss": 1.2038,
+      "step": 20
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 2.1000000000000002e-06,
+      "loss": 1.1958,
+      "step": 21
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 2.2e-06,
+      "loss": 0.9723,
+      "step": 22
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 2.3000000000000004e-06,
+      "loss": 1.0598,
+      "step": 23
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 2.4000000000000003e-06,
+      "loss": 0.8501,
+      "step": 24
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 2.5e-06,
+      "loss": 0.8728,
+      "step": 25
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 2.6e-06,
+      "loss": 1.1256,
+      "step": 26
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 2.7000000000000004e-06,
+      "loss": 0.9702,
+      "step": 27
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 2.8000000000000003e-06,
+      "loss": 1.1258,
+      "step": 28
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 2.9e-06,
+      "loss": 1.0507,
+      "step": 29
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 3e-06,
+      "loss": 0.9789,
+      "step": 30
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 3.1000000000000004e-06,
+      "loss": 0.8359,
+      "step": 31
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 3.2000000000000003e-06,
+      "loss": 0.9653,
+      "step": 32
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 3.3000000000000006e-06,
+      "loss": 0.8996,
+      "step": 33
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 3.4000000000000005e-06,
+      "loss": 1.0018,
+      "step": 34
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 3.5e-06,
+      "loss": 0.8889,
+      "step": 35
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 3.6000000000000003e-06,
+      "loss": 0.9844,
+      "step": 36
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 3.7e-06,
+      "loss": 0.9551,
+      "step": 37
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 3.8000000000000005e-06,
+      "loss": 0.8327,
+      "step": 38
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 3.900000000000001e-06,
+      "loss": 0.8914,
+      "step": 39
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 4.000000000000001e-06,
+      "loss": 0.8989,
+      "step": 40
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 4.1e-06,
+      "loss": 0.703,
+      "step": 41
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 4.2000000000000004e-06,
+      "loss": 0.884,
+      "step": 42
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 4.3e-06,
+      "loss": 0.7519,
+      "step": 43
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 4.4e-06,
+      "loss": 0.8547,
+      "step": 44
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 4.5e-06,
+      "loss": 0.8231,
+      "step": 45
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 4.600000000000001e-06,
+      "loss": 0.7351,
+      "step": 46
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 4.7e-06,
+      "loss": 0.7906,
+      "step": 47
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 4.800000000000001e-06,
+      "loss": 0.7365,
+      "step": 48
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 4.9000000000000005e-06,
+      "loss": 0.7611,
+      "step": 49
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 5e-06,
+      "loss": 0.6853,
+      "step": 50
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 5.1e-06,
+      "loss": 0.6458,
+      "step": 51
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 5.2e-06,
+      "loss": 0.712,
+      "step": 52
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 5.300000000000001e-06,
+      "loss": 0.6755,
+      "step": 53
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 5.400000000000001e-06,
+      "loss": 0.6172,
+      "step": 54
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 5.500000000000001e-06,
+      "loss": 0.6864,
+      "step": 55
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 5.600000000000001e-06,
+      "loss": 0.6047,
+      "step": 56
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 5.7e-06,
+      "loss": 0.5952,
+      "step": 57
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 5.8e-06,
+      "loss": 0.6368,
+      "step": 58
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 5.9e-06,
+      "loss": 0.5363,
+      "step": 59
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 6e-06,
+      "loss": 0.5781,
+      "step": 60
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 6.1e-06,
+      "loss": 0.6068,
+      "step": 61
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 6.200000000000001e-06,
+      "loss": 0.5856,
+      "step": 62
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 6.300000000000001e-06,
+      "loss": 0.5719,
+      "step": 63
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 6.4000000000000006e-06,
+      "loss": 0.5862,
+      "step": 64
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 6.5000000000000004e-06,
+      "loss": 0.5424,
+      "step": 65
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 6.600000000000001e-06,
+      "loss": 0.4901,
+      "step": 66
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 6.700000000000001e-06,
+      "loss": 0.4572,
+      "step": 67
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 6.800000000000001e-06,
+      "loss": 0.4476,
+      "step": 68
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 6.9e-06,
+      "loss": 0.5266,
+      "step": 69
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 7e-06,
+      "loss": 0.5051,
+      "step": 70
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 7.100000000000001e-06,
+      "loss": 0.4577,
+      "step": 71
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 7.2000000000000005e-06,
+      "loss": 0.4483,
+      "step": 72
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 7.3e-06,
+      "loss": 0.5268,
+      "step": 73
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 7.4e-06,
+      "loss": 0.4774,
+      "step": 74
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 7.500000000000001e-06,
+      "loss": 0.3889,
+      "step": 75
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 7.600000000000001e-06,
+      "loss": 0.5319,
+      "step": 76
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 7.7e-06,
+      "loss": 0.4035,
+      "step": 77
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 7.800000000000002e-06,
+      "loss": 0.5946,
+      "step": 78
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 7.9e-06,
+      "loss": 0.4474,
+      "step": 79
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 8.000000000000001e-06,
+      "loss": 0.4996,
+      "step": 80
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 8.1e-06,
+      "loss": 0.6067,
+      "step": 81
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 8.2e-06,
+      "loss": 0.4301,
+      "step": 82
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 8.3e-06,
+      "loss": 0.4119,
+      "step": 83
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 8.400000000000001e-06,
+      "loss": 0.4311,
+      "step": 84
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 8.5e-06,
+      "loss": 0.4479,
+      "step": 85
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 8.6e-06,
+      "loss": 0.3555,
+      "step": 86
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 8.700000000000001e-06,
+      "loss": 0.4877,
+      "step": 87
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 8.8e-06,
+      "loss": 0.3912,
+      "step": 88
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 8.900000000000001e-06,
+      "loss": 0.3578,
+      "step": 89
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 9e-06,
+      "loss": 0.3706,
+      "step": 90
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 9.100000000000001e-06,
+      "loss": 0.3889,
+      "step": 91
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 9.200000000000002e-06,
+      "loss": 0.4256,
+      "step": 92
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 9.3e-06,
+      "loss": 0.3372,
+      "step": 93
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 9.4e-06,
+      "loss": 0.4404,
+      "step": 94
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 9.5e-06,
+      "loss": 0.3325,
+      "step": 95
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 9.600000000000001e-06,
+      "loss": 0.4104,
+      "step": 96
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 9.7e-06,
+      "loss": 0.4069,
+      "step": 97
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 9.800000000000001e-06,
+      "loss": 0.3977,
+      "step": 98
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 9.9e-06,
+      "loss": 0.47,
+      "step": 99
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1e-05,
+      "loss": 0.3292,
+      "step": 100
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 9.999866702212744e-06,
+      "loss": 0.4137,
+      "step": 101
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 9.999733404425488e-06,
+      "loss": 0.3597,
+      "step": 102
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 9.999600106638231e-06,
+      "loss": 0.3905,
+      "step": 103
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 9.999466808850974e-06,
+      "loss": 0.3795,
+      "step": 104
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 9.999333511063717e-06,
+      "loss": 0.3564,
+      "step": 105
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 9.999200213276461e-06,
+      "loss": 0.4117,
+      "step": 106
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 9.999066915489204e-06,
+      "loss": 0.3322,
+      "step": 107
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 9.998933617701947e-06,
+      "loss": 0.3209,
+      "step": 108
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 9.99880031991469e-06,
+      "loss": 0.2899,
+      "step": 109
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 9.998667022127434e-06,
+      "loss": 0.3324,
+      "step": 110
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 9.998533724340177e-06,
+      "loss": 0.4101,
+      "step": 111
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 9.99840042655292e-06,
+      "loss": 0.4075,
+      "step": 112
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 9.998267128765663e-06,
+      "loss": 0.2901,
+      "step": 113
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 9.998133830978407e-06,
+      "loss": 0.3618,
+      "step": 114
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 9.99800053319115e-06,
+      "loss": 0.3754,
+      "step": 115
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 9.997867235403893e-06,
+      "loss": 0.2869,
+      "step": 116
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 9.997733937616636e-06,
+      "loss": 0.4456,
+      "step": 117
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 9.99760063982938e-06,
+      "loss": 0.3589,
+      "step": 118
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 9.997467342042123e-06,
+      "loss": 0.3046,
+      "step": 119
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 9.997334044254866e-06,
+      "loss": 0.3358,
+      "step": 120
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 9.997200746467608e-06,
+      "loss": 0.3456,
+      "step": 121
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 9.997067448680353e-06,
+      "loss": 0.3451,
+      "step": 122
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 9.996934150893096e-06,
+      "loss": 0.3562,
+      "step": 123
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 9.996800853105839e-06,
+      "loss": 0.301,
+      "step": 124
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 9.996667555318581e-06,
+      "loss": 0.379,
+      "step": 125
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.996534257531326e-06,
+      "loss": 0.3841,
+      "step": 126
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.996400959744069e-06,
+      "loss": 0.3027,
+      "step": 127
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.996267661956812e-06,
+      "loss": 0.3442,
+      "step": 128
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.996134364169554e-06,
+      "loss": 0.3396,
+      "step": 129
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.996001066382299e-06,
+      "loss": 0.3589,
+      "step": 130
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.995867768595042e-06,
+      "loss": 0.3797,
+      "step": 131
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.995734470807785e-06,
+      "loss": 0.3317,
+      "step": 132
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.995601173020527e-06,
+      "loss": 0.3804,
+      "step": 133
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.995467875233272e-06,
+      "loss": 0.3244,
+      "step": 134
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.995334577446015e-06,
+      "loss": 0.3035,
+      "step": 135
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.995201279658758e-06,
+      "loss": 0.3823,
+      "step": 136
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.995067981871502e-06,
+      "loss": 0.3087,
+      "step": 137
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.994934684084245e-06,
+      "loss": 0.3275,
+      "step": 138
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.994801386296988e-06,
+      "loss": 0.318,
+      "step": 139
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.99466808850973e-06,
+      "loss": 0.3507,
+      "step": 140
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.994534790722475e-06,
+      "loss": 0.3125,
+      "step": 141
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.994401492935218e-06,
+      "loss": 0.28,
+      "step": 142
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.99426819514796e-06,
+      "loss": 0.2708,
+      "step": 143
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.994134897360705e-06,
+      "loss": 0.2941,
+      "step": 144
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.994001599573448e-06,
+      "loss": 0.339,
+      "step": 145
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.99386830178619e-06,
+      "loss": 0.2508,
+      "step": 146
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.993735003998934e-06,
+      "loss": 0.2983,
+      "step": 147
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.993601706211678e-06,
+      "loss": 0.3023,
+      "step": 148
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.993468408424421e-06,
+      "loss": 0.3689,
+      "step": 149
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.993335110637164e-06,
+      "loss": 0.2918,
+      "step": 150
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.993201812849908e-06,
+      "loss": 0.3157,
+      "step": 151
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.993068515062651e-06,
+      "loss": 0.3471,
+      "step": 152
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.992935217275394e-06,
+      "loss": 0.2485,
+      "step": 153
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.992801919488137e-06,
+      "loss": 0.2765,
+      "step": 154
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.992668621700881e-06,
+      "loss": 0.2795,
+      "step": 155
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.992535323913624e-06,
+      "loss": 0.3035,
+      "step": 156
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.992402026126367e-06,
+      "loss": 0.3065,
+      "step": 157
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.992268728339111e-06,
+      "loss": 0.2798,
+      "step": 158
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.992135430551854e-06,
+      "loss": 0.2647,
+      "step": 159
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.992002132764597e-06,
+      "loss": 0.3191,
+      "step": 160
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.99186883497734e-06,
+      "loss": 0.2611,
+      "step": 161
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.991735537190084e-06,
+      "loss": 0.2834,
+      "step": 162
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.991602239402827e-06,
+      "loss": 0.3579,
+      "step": 163
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.99146894161557e-06,
+      "loss": 0.2927,
+      "step": 164
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.991335643828315e-06,
+      "loss": 0.3214,
+      "step": 165
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.991202346041057e-06,
+      "loss": 0.2955,
+      "step": 166
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.9910690482538e-06,
+      "loss": 0.304,
+      "step": 167
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.990935750466543e-06,
+      "loss": 0.3134,
+      "step": 168
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.990802452679288e-06,
+      "loss": 0.2993,
+      "step": 169
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.99066915489203e-06,
+      "loss": 0.2541,
+      "step": 170
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.990535857104773e-06,
+      "loss": 0.2284,
+      "step": 171
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.990402559317516e-06,
+      "loss": 0.2952,
+      "step": 172
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.99026926153026e-06,
+      "loss": 0.258,
+      "step": 173
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.990135963743003e-06,
+      "loss": 0.2711,
+      "step": 174
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.990002665955746e-06,
+      "loss": 0.2751,
+      "step": 175
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.989869368168489e-06,
+      "loss": 0.2949,
+      "step": 176
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.989736070381233e-06,
+      "loss": 0.2647,
+      "step": 177
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.989602772593976e-06,
+      "loss": 0.2864,
+      "step": 178
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.989469474806719e-06,
+      "loss": 0.2175,
+      "step": 179
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.989336177019462e-06,
+      "loss": 0.2628,
+      "step": 180
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.989202879232206e-06,
+      "loss": 0.3199,
+      "step": 181
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.98906958144495e-06,
+      "loss": 0.2391,
+      "step": 182
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.988936283657692e-06,
+      "loss": 0.2495,
+      "step": 183
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.988802985870435e-06,
+      "loss": 0.2704,
+      "step": 184
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.98866968808318e-06,
+      "loss": 0.2108,
+      "step": 185
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.988536390295922e-06,
+      "loss": 0.3196,
+      "step": 186
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.988403092508665e-06,
+      "loss": 0.276,
+      "step": 187
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.988269794721408e-06,
+      "loss": 0.3185,
+      "step": 188
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.988136496934152e-06,
+      "loss": 0.234,
+      "step": 189
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.988003199146895e-06,
+      "loss": 0.2377,
+      "step": 190
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.987869901359638e-06,
+      "loss": 0.2593,
+      "step": 191
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.98773660357238e-06,
+      "loss": 0.3055,
+      "step": 192
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.987603305785125e-06,
+      "loss": 0.2497,
+      "step": 193
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.987470007997868e-06,
+      "loss": 0.2585,
+      "step": 194
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.987336710210611e-06,
+      "loss": 0.2356,
+      "step": 195
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.987203412423354e-06,
+      "loss": 0.2526,
+      "step": 196
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.987070114636098e-06,
+      "loss": 0.306,
+      "step": 197
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.986936816848841e-06,
+      "loss": 0.2766,
+      "step": 198
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.986803519061584e-06,
+      "loss": 0.2694,
+      "step": 199
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.986670221274327e-06,
+      "loss": 0.2712,
+      "step": 200
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.986536923487071e-06,
+      "loss": 0.2296,
+      "step": 201
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.986403625699814e-06,
+      "loss": 0.2322,
+      "step": 202
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.986270327912557e-06,
+      "loss": 0.308,
+      "step": 203
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.9861370301253e-06,
+      "loss": 0.2734,
+      "step": 204
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.986003732338044e-06,
+      "loss": 0.373,
+      "step": 205
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.985870434550787e-06,
+      "loss": 0.2614,
+      "step": 206
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.98573713676353e-06,
+      "loss": 0.2577,
+      "step": 207
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.985603838976273e-06,
+      "loss": 0.2611,
+      "step": 208
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.985470541189017e-06,
+      "loss": 0.2841,
+      "step": 209
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.98533724340176e-06,
+      "loss": 0.2529,
+      "step": 210
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.985203945614503e-06,
+      "loss": 0.3231,
+      "step": 211
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.985070647827246e-06,
+      "loss": 0.2378,
+      "step": 212
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.98493735003999e-06,
+      "loss": 0.2943,
+      "step": 213
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.984804052252733e-06,
+      "loss": 0.2543,
+      "step": 214
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.984670754465476e-06,
+      "loss": 0.2852,
+      "step": 215
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.98453745667822e-06,
+      "loss": 0.2654,
+      "step": 216
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.984404158890963e-06,
+      "loss": 0.2501,
+      "step": 217
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.984270861103706e-06,
+      "loss": 0.2566,
+      "step": 218
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.984137563316449e-06,
+      "loss": 0.2683,
+      "step": 219
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.984004265529193e-06,
+      "loss": 0.2632,
+      "step": 220
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.983870967741936e-06,
+      "loss": 0.2654,
+      "step": 221
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.983737669954679e-06,
+      "loss": 0.2296,
+      "step": 222
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.983604372167422e-06,
+      "loss": 0.2424,
+      "step": 223
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.983471074380166e-06,
+      "loss": 0.2829,
+      "step": 224
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.983337776592909e-06,
+      "loss": 0.2626,
+      "step": 225
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.983204478805652e-06,
+      "loss": 0.2232,
+      "step": 226
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.983071181018395e-06,
+      "loss": 0.2274,
+      "step": 227
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.98293788323114e-06,
+      "loss": 0.314,
+      "step": 228
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.982804585443882e-06,
+      "loss": 0.2423,
+      "step": 229
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.982671287656625e-06,
+      "loss": 0.2245,
+      "step": 230
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.98253798986937e-06,
+      "loss": 0.2669,
+      "step": 231
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.982404692082112e-06,
+      "loss": 0.2575,
+      "step": 232
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.982271394294855e-06,
+      "loss": 0.2434,
+      "step": 233
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.982138096507598e-06,
+      "loss": 0.2384,
+      "step": 234
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.982004798720342e-06,
+      "loss": 0.2143,
+      "step": 235
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.981871500933085e-06,
+      "loss": 0.2146,
+      "step": 236
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.981738203145828e-06,
+      "loss": 0.2148,
+      "step": 237
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.981604905358573e-06,
+      "loss": 0.2109,
+      "step": 238
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.981471607571315e-06,
+      "loss": 0.207,
+      "step": 239
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.981338309784058e-06,
+      "loss": 0.2262,
+      "step": 240
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.981205011996801e-06,
+      "loss": 0.2663,
+      "step": 241
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.981071714209545e-06,
+      "loss": 0.2131,
+      "step": 242
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.980938416422288e-06,
+      "loss": 0.2106,
+      "step": 243
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.980805118635031e-06,
+      "loss": 0.2212,
+      "step": 244
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.980671820847776e-06,
+      "loss": 0.2108,
+      "step": 245
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.980538523060518e-06,
+      "loss": 0.3081,
+      "step": 246
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.980405225273261e-06,
+      "loss": 0.2827,
+      "step": 247
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.980271927486004e-06,
+      "loss": 0.2791,
+      "step": 248
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.980138629698749e-06,
+      "loss": 0.2734,
+      "step": 249
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.980005331911491e-06,
+      "loss": 0.244,
+      "step": 250
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.979872034124234e-06,
+      "loss": 0.2329,
+      "step": 251
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.979738736336979e-06,
+      "loss": 0.2721,
+      "step": 252
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.979605438549722e-06,
+      "loss": 0.1978,
+      "step": 253
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.979472140762464e-06,
+      "loss": 0.2301,
+      "step": 254
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.979338842975207e-06,
+      "loss": 0.171,
+      "step": 255
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.979205545187952e-06,
+      "loss": 0.2484,
+      "step": 256
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.979072247400695e-06,
+      "loss": 0.1941,
+      "step": 257
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.978938949613437e-06,
+      "loss": 0.298,
+      "step": 258
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.97880565182618e-06,
+      "loss": 0.2461,
+      "step": 259
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.978672354038925e-06,
+      "loss": 0.2526,
+      "step": 260
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.978539056251668e-06,
+      "loss": 0.1747,
+      "step": 261
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.97840575846441e-06,
+      "loss": 0.2062,
+      "step": 262
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.978272460677153e-06,
+      "loss": 0.2439,
+      "step": 263
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.978139162889898e-06,
+      "loss": 0.273,
+      "step": 264
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.97800586510264e-06,
+      "loss": 0.2838,
+      "step": 265
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.977872567315383e-06,
+      "loss": 0.217,
+      "step": 266
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.977739269528126e-06,
+      "loss": 0.2484,
+      "step": 267
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.97760597174087e-06,
+      "loss": 0.2254,
+      "step": 268
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.977472673953613e-06,
+      "loss": 0.2451,
+      "step": 269
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.977339376166356e-06,
+      "loss": 0.2048,
+      "step": 270
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.9772060783791e-06,
+      "loss": 0.3033,
+      "step": 271
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.977072780591844e-06,
+      "loss": 0.2393,
+      "step": 272
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.976939482804586e-06,
+      "loss": 0.2174,
+      "step": 273
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.97680618501733e-06,
+      "loss": 0.2839,
+      "step": 274
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.976672887230074e-06,
+      "loss": 0.2567,
+      "step": 275
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.976539589442817e-06,
+      "loss": 0.2696,
+      "step": 276
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.97640629165556e-06,
+      "loss": 0.265,
+      "step": 277
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.976272993868302e-06,
+      "loss": 0.2521,
+      "step": 278
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.976139696081047e-06,
+      "loss": 0.2795,
+      "step": 279
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.97600639829379e-06,
+      "loss": 0.2528,
+      "step": 280
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.975873100506532e-06,
+      "loss": 0.2622,
+      "step": 281
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.975739802719275e-06,
+      "loss": 0.2223,
+      "step": 282
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.97560650493202e-06,
+      "loss": 0.1791,
+      "step": 283
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.975473207144763e-06,
+      "loss": 0.1852,
+      "step": 284
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.975339909357505e-06,
+      "loss": 0.208,
+      "step": 285
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.975206611570248e-06,
+      "loss": 0.2116,
+      "step": 286
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.975073313782993e-06,
+      "loss": 0.2073,
+      "step": 287
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.974940015995735e-06,
+      "loss": 0.2499,
+      "step": 288
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.974806718208478e-06,
+      "loss": 0.2036,
+      "step": 289
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.974673420421221e-06,
+      "loss": 0.2002,
+      "step": 290
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.974540122633966e-06,
+      "loss": 0.2174,
+      "step": 291
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.974406824846708e-06,
+      "loss": 0.2696,
+      "step": 292
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.974273527059451e-06,
+      "loss": 0.2361,
+      "step": 293
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.974140229272194e-06,
+      "loss": 0.2007,
+      "step": 294
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.974006931484939e-06,
+      "loss": 0.2188,
+      "step": 295
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.973873633697681e-06,
+      "loss": 0.2078,
+      "step": 296
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.973740335910424e-06,
+      "loss": 0.2195,
+      "step": 297
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.973607038123167e-06,
+      "loss": 0.1892,
+      "step": 298
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.973473740335912e-06,
+      "loss": 0.2222,
+      "step": 299
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.973340442548654e-06,
+      "loss": 0.2315,
+      "step": 300
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.973207144761397e-06,
+      "loss": 0.2185,
+      "step": 301
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.97307384697414e-06,
+      "loss": 0.2148,
+      "step": 302
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.972940549186885e-06,
+      "loss": 0.2228,
+      "step": 303
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.972807251399627e-06,
+      "loss": 0.1759,
+      "step": 304
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.97267395361237e-06,
+      "loss": 0.1981,
+      "step": 305
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.972540655825113e-06,
+      "loss": 0.2366,
+      "step": 306
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.972407358037858e-06,
+      "loss": 0.2247,
+      "step": 307
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.9722740602506e-06,
+      "loss": 0.2328,
+      "step": 308
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.972140762463343e-06,
+      "loss": 0.2712,
+      "step": 309
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.972007464676086e-06,
+      "loss": 0.2111,
+      "step": 310
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.97187416688883e-06,
+      "loss": 0.2068,
+      "step": 311
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.971740869101573e-06,
+      "loss": 0.1966,
+      "step": 312
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.971607571314316e-06,
+      "loss": 0.2331,
+      "step": 313
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.971474273527059e-06,
+      "loss": 0.1854,
+      "step": 314
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.971340975739803e-06,
+      "loss": 0.2479,
+      "step": 315
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.971207677952546e-06,
+      "loss": 0.2219,
+      "step": 316
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.971074380165289e-06,
+      "loss": 0.1735,
+      "step": 317
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.970941082378034e-06,
+      "loss": 0.2171,
+      "step": 318
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.970807784590776e-06,
+      "loss": 0.1955,
+      "step": 319
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.97067448680352e-06,
+      "loss": 0.1766,
+      "step": 320
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.970541189016262e-06,
+      "loss": 0.2598,
+      "step": 321
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.970407891229007e-06,
+      "loss": 0.1983,
+      "step": 322
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.97027459344175e-06,
+      "loss": 0.2088,
+      "step": 323
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.970141295654492e-06,
+      "loss": 0.2033,
+      "step": 324
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.970007997867237e-06,
+      "loss": 0.2359,
+      "step": 325
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.96987470007998e-06,
+      "loss": 0.2355,
+      "step": 326
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.969741402292722e-06,
+      "loss": 0.2323,
+      "step": 327
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.969608104505465e-06,
+      "loss": 0.2265,
+      "step": 328
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.96947480671821e-06,
+      "loss": 0.2111,
+      "step": 329
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.969341508930953e-06,
+      "loss": 0.2175,
+      "step": 330
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.969208211143695e-06,
+      "loss": 0.2193,
+      "step": 331
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.96907491335644e-06,
+      "loss": 0.2515,
+      "step": 332
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.968941615569183e-06,
+      "loss": 0.1859,
+      "step": 333
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.968808317781925e-06,
+      "loss": 0.236,
+      "step": 334
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.968675019994668e-06,
+      "loss": 0.2299,
+      "step": 335
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.968541722207413e-06,
+      "loss": 0.2232,
+      "step": 336
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.968408424420156e-06,
+      "loss": 0.1819,
+      "step": 337
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.968275126632898e-06,
+      "loss": 0.1866,
+      "step": 338
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.968141828845643e-06,
+      "loss": 0.2574,
+      "step": 339
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.968008531058386e-06,
+      "loss": 0.2325,
+      "step": 340
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.967875233271129e-06,
+      "loss": 0.2212,
+      "step": 341
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.967741935483871e-06,
+      "loss": 0.2192,
+      "step": 342
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.967608637696616e-06,
+      "loss": 0.1918,
+      "step": 343
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.967475339909359e-06,
+      "loss": 0.2243,
+      "step": 344
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.967342042122102e-06,
+      "loss": 0.2391,
+      "step": 345
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.967208744334846e-06,
+      "loss": 0.2287,
+      "step": 346
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.967075446547589e-06,
+      "loss": 0.192,
+      "step": 347
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.966942148760332e-06,
+      "loss": 0.2215,
+      "step": 348
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.966808850973075e-06,
+      "loss": 0.1705,
+      "step": 349
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.966675553185819e-06,
+      "loss": 0.1775,
+      "step": 350
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.966542255398562e-06,
+      "loss": 0.2048,
+      "step": 351
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.966408957611305e-06,
+      "loss": 0.2141,
+      "step": 352
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.966275659824048e-06,
+      "loss": 0.2265,
+      "step": 353
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.966142362036792e-06,
+      "loss": 0.2002,
+      "step": 354
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.966009064249535e-06,
+      "loss": 0.2152,
+      "step": 355
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.965875766462278e-06,
+      "loss": 0.1991,
+      "step": 356
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.96574246867502e-06,
+      "loss": 0.1985,
+      "step": 357
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.965609170887765e-06,
+      "loss": 0.2076,
+      "step": 358
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.965475873100508e-06,
+      "loss": 0.2167,
+      "step": 359
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.96534257531325e-06,
+      "loss": 0.239,
+      "step": 360
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.965209277525993e-06,
+      "loss": 0.1731,
+      "step": 361
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.965075979738738e-06,
+      "loss": 0.2006,
+      "step": 362
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.96494268195148e-06,
+      "loss": 0.1882,
+      "step": 363
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.964809384164224e-06,
+      "loss": 0.2155,
+      "step": 364
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.964676086376966e-06,
+      "loss": 0.1793,
+      "step": 365
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.964542788589711e-06,
+      "loss": 0.2205,
+      "step": 366
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.964409490802454e-06,
+      "loss": 0.1717,
+      "step": 367
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.964276193015197e-06,
+      "loss": 0.1876,
+      "step": 368
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.96414289522794e-06,
+      "loss": 0.1951,
+      "step": 369
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.964009597440684e-06,
+      "loss": 0.1272,
+      "step": 370
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.963876299653427e-06,
+      "loss": 0.1715,
+      "step": 371
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.96374300186617e-06,
+      "loss": 0.1494,
+      "step": 372
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.963609704078912e-06,
+      "loss": 0.1957,
+      "step": 373
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.963476406291657e-06,
+      "loss": 0.2448,
+      "step": 374
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.9633431085044e-06,
+      "loss": 0.1905,
+      "step": 375
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.963209810717143e-06,
+      "loss": 0.2125,
+      "step": 376
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.963076512929885e-06,
+      "loss": 0.2128,
+      "step": 377
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.96294321514263e-06,
+      "loss": 0.2136,
+      "step": 378
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.962809917355373e-06,
+      "loss": 0.216,
+      "step": 379
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.962676619568116e-06,
+      "loss": 0.1928,
+      "step": 380
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.962543321780858e-06,
+      "loss": 0.2432,
+      "step": 381
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.962410023993603e-06,
+      "loss": 0.1724,
+      "step": 382
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.962276726206346e-06,
+      "loss": 0.2409,
+      "step": 383
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.962143428419088e-06,
+      "loss": 0.1632,
+      "step": 384
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.962010130631833e-06,
+      "loss": 0.1845,
+      "step": 385
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.961876832844576e-06,
+      "loss": 0.2127,
+      "step": 386
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.961743535057319e-06,
+      "loss": 0.2163,
+      "step": 387
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.961610237270061e-06,
+      "loss": 0.2409,
+      "step": 388
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.961476939482806e-06,
+      "loss": 0.2279,
+      "step": 389
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.961343641695549e-06,
+      "loss": 0.2303,
+      "step": 390
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.961210343908292e-06,
+      "loss": 0.2285,
+      "step": 391
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.961077046121034e-06,
+      "loss": 0.1884,
+      "step": 392
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.960943748333779e-06,
+      "loss": 0.1826,
+      "step": 393
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.960810450546522e-06,
+      "loss": 0.192,
+      "step": 394
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.960677152759265e-06,
+      "loss": 0.2114,
+      "step": 395
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.960543854972007e-06,
+      "loss": 0.1949,
+      "step": 396
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.960410557184752e-06,
+      "loss": 0.1871,
+      "step": 397
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.960277259397495e-06,
+      "loss": 0.1848,
+      "step": 398
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.960143961610238e-06,
+      "loss": 0.1901,
+      "step": 399
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.96001066382298e-06,
+      "loss": 0.2432,
+      "step": 400
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.959877366035725e-06,
+      "loss": 0.2295,
+      "step": 401
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.959744068248468e-06,
+      "loss": 0.2045,
+      "step": 402
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.95961077046121e-06,
+      "loss": 0.2009,
+      "step": 403
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.959477472673953e-06,
+      "loss": 0.2044,
+      "step": 404
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.959344174886698e-06,
+      "loss": 0.2251,
+      "step": 405
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.95921087709944e-06,
+      "loss": 0.2486,
+      "step": 406
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.959077579312183e-06,
+      "loss": 0.1897,
+      "step": 407
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.958944281524926e-06,
+      "loss": 0.2208,
+      "step": 408
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.95881098373767e-06,
+      "loss": 0.2308,
+      "step": 409
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.958677685950414e-06,
+      "loss": 0.2256,
+      "step": 410
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.958544388163156e-06,
+      "loss": 0.1759,
+      "step": 411
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.958411090375901e-06,
+      "loss": 0.2234,
+      "step": 412
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.958277792588644e-06,
+      "loss": 0.2309,
+      "step": 413
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.958144494801387e-06,
+      "loss": 0.1627,
+      "step": 414
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.95801119701413e-06,
+      "loss": 0.1762,
+      "step": 415
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.957877899226874e-06,
+      "loss": 0.2459,
+      "step": 416
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.957744601439617e-06,
+      "loss": 0.2292,
+      "step": 417
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.95761130365236e-06,
+      "loss": 0.1528,
+      "step": 418
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.957478005865104e-06,
+      "loss": 0.204,
+      "step": 419
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.957344708077847e-06,
+      "loss": 0.2003,
+      "step": 420
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.95721141029059e-06,
+      "loss": 0.166,
+      "step": 421
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.957078112503333e-06,
+      "loss": 0.1674,
+      "step": 422
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.956944814716077e-06,
+      "loss": 0.1642,
+      "step": 423
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.95681151692882e-06,
+      "loss": 0.1706,
+      "step": 424
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.956678219141563e-06,
+      "loss": 0.1557,
+      "step": 425
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.956544921354307e-06,
+      "loss": 0.1716,
+      "step": 426
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.95641162356705e-06,
+      "loss": 0.1799,
+      "step": 427
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.956278325779793e-06,
+      "loss": 0.1879,
+      "step": 428
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.956145027992536e-06,
+      "loss": 0.2326,
+      "step": 429
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.95601173020528e-06,
+      "loss": 0.2191,
+      "step": 430
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.955878432418023e-06,
+      "loss": 0.2121,
+      "step": 431
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.955745134630766e-06,
+      "loss": 0.2485,
+      "step": 432
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.95561183684351e-06,
+      "loss": 0.2327,
+      "step": 433
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.955478539056253e-06,
+      "loss": 0.1954,
+      "step": 434
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.955345241268996e-06,
+      "loss": 0.1432,
+      "step": 435
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.955211943481739e-06,
+      "loss": 0.2018,
+      "step": 436
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.955078645694483e-06,
+      "loss": 0.1769,
+      "step": 437
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.954945347907226e-06,
+      "loss": 0.205,
+      "step": 438
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.954812050119969e-06,
+      "loss": 0.1834,
+      "step": 439
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.954678752332713e-06,
+      "loss": 0.1641,
+      "step": 440
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.954545454545456e-06,
+      "loss": 0.2168,
+      "step": 441
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.954412156758199e-06,
+      "loss": 0.2191,
+      "step": 442
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.954278858970942e-06,
+      "loss": 0.236,
+      "step": 443
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.954145561183686e-06,
+      "loss": 0.2504,
+      "step": 444
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.95401226339643e-06,
+      "loss": 0.1767,
+      "step": 445
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.953878965609172e-06,
+      "loss": 0.2016,
+      "step": 446
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.953745667821915e-06,
+      "loss": 0.139,
+      "step": 447
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.95361237003466e-06,
+      "loss": 0.1341,
+      "step": 448
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.953479072247402e-06,
+      "loss": 0.2284,
+      "step": 449
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.953345774460145e-06,
+      "loss": 0.168,
+      "step": 450
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.953212476672888e-06,
+      "loss": 0.1672,
+      "step": 451
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.953079178885632e-06,
+      "loss": 0.238,
+      "step": 452
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.952945881098375e-06,
+      "loss": 0.1706,
+      "step": 453
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.952812583311118e-06,
+      "loss": 0.1689,
+      "step": 454
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.95267928552386e-06,
+      "loss": 0.2128,
+      "step": 455
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.952545987736605e-06,
+      "loss": 0.2185,
+      "step": 456
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.952412689949348e-06,
+      "loss": 0.1851,
+      "step": 457
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.952279392162091e-06,
+      "loss": 0.1793,
+      "step": 458
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.952146094374834e-06,
+      "loss": 0.1969,
+      "step": 459
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.952012796587578e-06,
+      "loss": 0.17,
+      "step": 460
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.951879498800321e-06,
+      "loss": 0.1812,
+      "step": 461
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.951746201013064e-06,
+      "loss": 0.1882,
+      "step": 462
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.951612903225807e-06,
+      "loss": 0.1678,
+      "step": 463
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.951479605438551e-06,
+      "loss": 0.2108,
+      "step": 464
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.951346307651294e-06,
+      "loss": 0.1847,
+      "step": 465
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.951213009864037e-06,
+      "loss": 0.1964,
+      "step": 466
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.95107971207678e-06,
+      "loss": 0.2409,
+      "step": 467
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.950946414289524e-06,
+      "loss": 0.1728,
+      "step": 468
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.950813116502267e-06,
+      "loss": 0.1888,
+      "step": 469
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.95067981871501e-06,
+      "loss": 0.201,
+      "step": 470
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.950546520927753e-06,
+      "loss": 0.1965,
+      "step": 471
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.950413223140497e-06,
+      "loss": 0.1726,
+      "step": 472
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.95027992535324e-06,
+      "loss": 0.1448,
+      "step": 473
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.950146627565983e-06,
+      "loss": 0.1819,
+      "step": 474
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.950013329778726e-06,
+      "loss": 0.1928,
+      "step": 475
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.94988003199147e-06,
+      "loss": 0.1982,
+      "step": 476
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.949746734204213e-06,
+      "loss": 0.2022,
+      "step": 477
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.949613436416956e-06,
+      "loss": 0.1753,
+      "step": 478
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.949480138629699e-06,
+      "loss": 0.2047,
+      "step": 479
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.949346840842443e-06,
+      "loss": 0.1592,
+      "step": 480
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.949213543055186e-06,
+      "loss": 0.2255,
+      "step": 481
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.949080245267929e-06,
+      "loss": 0.1797,
+      "step": 482
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.948946947480672e-06,
+      "loss": 0.2411,
+      "step": 483
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.948813649693416e-06,
+      "loss": 0.2061,
+      "step": 484
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.948680351906159e-06,
+      "loss": 0.2177,
+      "step": 485
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.948547054118902e-06,
+      "loss": 0.2384,
+      "step": 486
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.948413756331645e-06,
+      "loss": 0.1846,
+      "step": 487
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.948280458544389e-06,
+      "loss": 0.241,
+      "step": 488
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.948147160757132e-06,
+      "loss": 0.2319,
+      "step": 489
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.948013862969875e-06,
+      "loss": 0.1587,
+      "step": 490
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.947880565182618e-06,
+      "loss": 0.1702,
+      "step": 491
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.947747267395362e-06,
+      "loss": 0.183,
+      "step": 492
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.947613969608105e-06,
+      "loss": 0.1297,
+      "step": 493
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.947480671820848e-06,
+      "loss": 0.2208,
+      "step": 494
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.947347374033592e-06,
+      "loss": 0.2033,
+      "step": 495
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.947214076246335e-06,
+      "loss": 0.179,
+      "step": 496
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.947080778459078e-06,
+      "loss": 0.204,
+      "step": 497
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.94694748067182e-06,
+      "loss": 0.1981,
+      "step": 498
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.946814182884565e-06,
+      "loss": 0.175,
+      "step": 499
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.946680885097308e-06,
+      "loss": 0.239,
+      "step": 500
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.94654758731005e-06,
+      "loss": 0.1766,
+      "step": 501
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.946414289522794e-06,
+      "loss": 0.2208,
+      "step": 502
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.946280991735538e-06,
+      "loss": 0.1757,
+      "step": 503
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.946147693948281e-06,
+      "loss": 0.1888,
+      "step": 504
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.946014396161024e-06,
+      "loss": 0.1662,
+      "step": 505
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.945881098373768e-06,
+      "loss": 0.167,
+      "step": 506
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.945747800586511e-06,
+      "loss": 0.179,
+      "step": 507
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.945614502799254e-06,
+      "loss": 0.2478,
+      "step": 508
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.945481205011997e-06,
+      "loss": 0.1778,
+      "step": 509
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.945347907224741e-06,
+      "loss": 0.1596,
+      "step": 510
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.945214609437484e-06,
+      "loss": 0.1743,
+      "step": 511
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.945081311650227e-06,
+      "loss": 0.1909,
+      "step": 512
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.944948013862971e-06,
+      "loss": 0.2093,
+      "step": 513
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.944814716075714e-06,
+      "loss": 0.2196,
+      "step": 514
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.944681418288457e-06,
+      "loss": 0.1832,
+      "step": 515
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.9445481205012e-06,
+      "loss": 0.1884,
+      "step": 516
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.944414822713944e-06,
+      "loss": 0.1987,
+      "step": 517
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.944281524926687e-06,
+      "loss": 0.1826,
+      "step": 518
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.94414822713943e-06,
+      "loss": 0.2217,
+      "step": 519
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.944014929352175e-06,
+      "loss": 0.1992,
+      "step": 520
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.943881631564917e-06,
+      "loss": 0.1945,
+      "step": 521
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.94374833377766e-06,
+      "loss": 0.1686,
+      "step": 522
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.943615035990403e-06,
+      "loss": 0.1931,
+      "step": 523
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.943481738203148e-06,
+      "loss": 0.1533,
+      "step": 524
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.94334844041589e-06,
+      "loss": 0.211,
+      "step": 525
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.943215142628633e-06,
+      "loss": 0.2226,
+      "step": 526
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.943081844841378e-06,
+      "loss": 0.1583,
+      "step": 527
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.94294854705412e-06,
+      "loss": 0.2328,
+      "step": 528
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.942815249266863e-06,
+      "loss": 0.2064,
+      "step": 529
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.942681951479606e-06,
+      "loss": 0.2281,
+      "step": 530
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.94254865369235e-06,
+      "loss": 0.1782,
+      "step": 531
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.942415355905093e-06,
+      "loss": 0.1613,
+      "step": 532
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.942282058117836e-06,
+      "loss": 0.1661,
+      "step": 533
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.942148760330579e-06,
+      "loss": 0.1514,
+      "step": 534
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.942015462543324e-06,
+      "loss": 0.2058,
+      "step": 535
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.941882164756066e-06,
+      "loss": 0.2076,
+      "step": 536
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.94174886696881e-06,
+      "loss": 0.1974,
+      "step": 537
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.941615569181552e-06,
+      "loss": 0.2201,
+      "step": 538
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.941482271394297e-06,
+      "loss": 0.1856,
+      "step": 539
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.94134897360704e-06,
+      "loss": 0.2188,
+      "step": 540
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.941215675819782e-06,
+      "loss": 0.1779,
+      "step": 541
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.941082378032525e-06,
+      "loss": 0.1967,
+      "step": 542
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.94094908024527e-06,
+      "loss": 0.1733,
+      "step": 543
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.940815782458012e-06,
+      "loss": 0.1181,
+      "step": 544
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.940682484670755e-06,
+      "loss": 0.2046,
+      "step": 545
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.940549186883498e-06,
+      "loss": 0.1776,
+      "step": 546
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.940415889096243e-06,
+      "loss": 0.1466,
+      "step": 547
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.940282591308985e-06,
+      "loss": 0.2233,
+      "step": 548
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.940149293521728e-06,
+      "loss": 0.1821,
+      "step": 549
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.940015995734471e-06,
+      "loss": 0.1532,
+      "step": 550
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.939882697947215e-06,
+      "loss": 0.1437,
+      "step": 551
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.939749400159958e-06,
+      "loss": 0.1371,
+      "step": 552
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.939616102372701e-06,
+      "loss": 0.1858,
+      "step": 553
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.939482804585446e-06,
+      "loss": 0.1922,
+      "step": 554
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.939349506798188e-06,
+      "loss": 0.1766,
+      "step": 555
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.939216209010931e-06,
+      "loss": 0.1887,
+      "step": 556
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.939082911223674e-06,
+      "loss": 0.1168,
+      "step": 557
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.938949613436419e-06,
+      "loss": 0.2044,
+      "step": 558
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.938816315649161e-06,
+      "loss": 0.1898,
+      "step": 559
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.938683017861904e-06,
+      "loss": 0.1608,
+      "step": 560
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.938549720074647e-06,
+      "loss": 0.1969,
+      "step": 561
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.938416422287392e-06,
+      "loss": 0.2167,
+      "step": 562
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.938283124500134e-06,
+      "loss": 0.1404,
+      "step": 563
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.938149826712877e-06,
+      "loss": 0.1834,
+      "step": 564
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.93801652892562e-06,
+      "loss": 0.183,
+      "step": 565
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.937883231138365e-06,
+      "loss": 0.1832,
+      "step": 566
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.937749933351107e-06,
+      "loss": 0.1732,
+      "step": 567
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.93761663556385e-06,
+      "loss": 0.1467,
+      "step": 568
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.937483337776593e-06,
+      "loss": 0.1815,
+      "step": 569
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.937350039989338e-06,
+      "loss": 0.2087,
+      "step": 570
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.93721674220208e-06,
+      "loss": 0.1751,
+      "step": 571
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.937083444414823e-06,
+      "loss": 0.1982,
+      "step": 572
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.936950146627566e-06,
+      "loss": 0.1867,
+      "step": 573
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.93681684884031e-06,
+      "loss": 0.2222,
+      "step": 574
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.936683551053053e-06,
+      "loss": 0.1978,
+      "step": 575
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.936550253265796e-06,
+      "loss": 0.1563,
+      "step": 576
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.936416955478539e-06,
+      "loss": 0.1833,
+      "step": 577
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.936283657691283e-06,
+      "loss": 0.1777,
+      "step": 578
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.936150359904026e-06,
+      "loss": 0.1659,
+      "step": 579
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.936017062116769e-06,
+      "loss": 0.1671,
+      "step": 580
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.935883764329512e-06,
+      "loss": 0.1773,
+      "step": 581
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.935750466542256e-06,
+      "loss": 0.1646,
+      "step": 582
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.935617168755e-06,
+      "loss": 0.1771,
+      "step": 583
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.935483870967742e-06,
+      "loss": 0.2103,
+      "step": 584
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.935350573180485e-06,
+      "loss": 0.1426,
+      "step": 585
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.93521727539323e-06,
+      "loss": 0.2256,
+      "step": 586
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.935083977605972e-06,
+      "loss": 0.1647,
+      "step": 587
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.934950679818715e-06,
+      "loss": 0.1421,
+      "step": 588
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.934817382031458e-06,
+      "loss": 0.1872,
+      "step": 589
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.934684084244202e-06,
+      "loss": 0.2479,
+      "step": 590
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.934550786456945e-06,
+      "loss": 0.1702,
+      "step": 591
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.934417488669688e-06,
+      "loss": 0.1582,
+      "step": 592
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.934284190882433e-06,
+      "loss": 0.1654,
+      "step": 593
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.934150893095175e-06,
+      "loss": 0.1743,
+      "step": 594
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.934017595307918e-06,
+      "loss": 0.1397,
+      "step": 595
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.933884297520661e-06,
+      "loss": 0.1599,
+      "step": 596
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.933750999733405e-06,
+      "loss": 0.2328,
+      "step": 597
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.933617701946148e-06,
+      "loss": 0.2086,
+      "step": 598
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.933484404158891e-06,
+      "loss": 0.154,
+      "step": 599
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.933351106371636e-06,
+      "loss": 0.155,
+      "step": 600
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.933217808584378e-06,
+      "loss": 0.1941,
+      "step": 601
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.933084510797121e-06,
+      "loss": 0.1988,
+      "step": 602
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.932951213009864e-06,
+      "loss": 0.1939,
+      "step": 603
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.932817915222609e-06,
+      "loss": 0.2051,
+      "step": 604
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.932684617435351e-06,
+      "loss": 0.1781,
+      "step": 605
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.932551319648094e-06,
+      "loss": 0.1732,
+      "step": 606
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.932418021860839e-06,
+      "loss": 0.1651,
+      "step": 607
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.932284724073582e-06,
+      "loss": 0.1776,
+      "step": 608
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.932151426286324e-06,
+      "loss": 0.1871,
+      "step": 609
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.932018128499067e-06,
+      "loss": 0.1868,
+      "step": 610
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.931884830711812e-06,
+      "loss": 0.2196,
+      "step": 611
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.931751532924555e-06,
+      "loss": 0.198,
+      "step": 612
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.931618235137297e-06,
+      "loss": 0.1652,
+      "step": 613
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.931484937350042e-06,
+      "loss": 0.1611,
+      "step": 614
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.931351639562785e-06,
+      "loss": 0.1824,
+      "step": 615
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.931218341775528e-06,
+      "loss": 0.2027,
+      "step": 616
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.93108504398827e-06,
+      "loss": 0.1958,
+      "step": 617
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.930951746201015e-06,
+      "loss": 0.237,
+      "step": 618
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.930818448413758e-06,
+      "loss": 0.1859,
+      "step": 619
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.9306851506265e-06,
+      "loss": 0.1719,
+      "step": 620
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.930551852839245e-06,
+      "loss": 0.2071,
+      "step": 621
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.930418555051988e-06,
+      "loss": 0.1644,
+      "step": 622
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.93028525726473e-06,
+      "loss": 0.1656,
+      "step": 623
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.930151959477473e-06,
+      "loss": 0.1571,
+      "step": 624
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.930018661690218e-06,
+      "loss": 0.1605,
+      "step": 625
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.92988536390296e-06,
+      "loss": 0.1395,
+      "step": 626
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.929752066115704e-06,
+      "loss": 0.1533,
+      "step": 627
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.929618768328446e-06,
+      "loss": 0.1287,
+      "step": 628
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.929485470541191e-06,
+      "loss": 0.2039,
+      "step": 629
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.929352172753934e-06,
+      "loss": 0.22,
+      "step": 630
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.929218874966677e-06,
+      "loss": 0.1955,
+      "step": 631
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.92908557717942e-06,
+      "loss": 0.1811,
+      "step": 632
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.928952279392164e-06,
+      "loss": 0.2012,
+      "step": 633
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.928818981604907e-06,
+      "loss": 0.1704,
+      "step": 634
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.92868568381765e-06,
+      "loss": 0.1996,
+      "step": 635
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.928552386030392e-06,
+      "loss": 0.2665,
+      "step": 636
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.928419088243137e-06,
+      "loss": 0.2115,
+      "step": 637
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.92828579045588e-06,
+      "loss": 0.1573,
+      "step": 638
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.928152492668623e-06,
+      "loss": 0.1969,
+      "step": 639
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.928019194881365e-06,
+      "loss": 0.2111,
+      "step": 640
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.92788589709411e-06,
+      "loss": 0.1911,
+      "step": 641
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.927752599306853e-06,
+      "loss": 0.1601,
+      "step": 642
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.927619301519595e-06,
+      "loss": 0.163,
+      "step": 643
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.927486003732338e-06,
+      "loss": 0.1691,
+      "step": 644
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.927352705945083e-06,
+      "loss": 0.1781,
+      "step": 645
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.927219408157826e-06,
+      "loss": 0.166,
+      "step": 646
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.927086110370568e-06,
+      "loss": 0.1948,
+      "step": 647
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.926952812583311e-06,
+      "loss": 0.1901,
+      "step": 648
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.926819514796056e-06,
+      "loss": 0.2003,
+      "step": 649
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.926686217008799e-06,
+      "loss": 0.1498,
+      "step": 650
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.926552919221541e-06,
+      "loss": 0.2387,
+      "step": 651
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.926419621434284e-06,
+      "loss": 0.1618,
+      "step": 652
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.926286323647029e-06,
+      "loss": 0.2223,
+      "step": 653
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.926153025859772e-06,
+      "loss": 0.1542,
+      "step": 654
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.926019728072514e-06,
+      "loss": 0.1616,
+      "step": 655
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.925886430285257e-06,
+      "loss": 0.1934,
+      "step": 656
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.925753132498002e-06,
+      "loss": 0.1437,
+      "step": 657
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.925619834710745e-06,
+      "loss": 0.172,
+      "step": 658
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.925486536923487e-06,
+      "loss": 0.2211,
+      "step": 659
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.92535323913623e-06,
+      "loss": 0.1705,
+      "step": 660
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.925219941348975e-06,
+      "loss": 0.2009,
+      "step": 661
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.925086643561718e-06,
+      "loss": 0.2043,
+      "step": 662
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.92495334577446e-06,
+      "loss": 0.1469,
+      "step": 663
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.924820047987205e-06,
+      "loss": 0.1422,
+      "step": 664
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.924686750199948e-06,
+      "loss": 0.1584,
+      "step": 665
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.92455345241269e-06,
+      "loss": 0.1479,
+      "step": 666
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.924420154625433e-06,
+      "loss": 0.1627,
+      "step": 667
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.924286856838178e-06,
+      "loss": 0.165,
+      "step": 668
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.92415355905092e-06,
+      "loss": 0.1532,
+      "step": 669
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.924020261263663e-06,
+      "loss": 0.1558,
+      "step": 670
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.923886963476406e-06,
+      "loss": 0.1519,
+      "step": 671
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.92375366568915e-06,
+      "loss": 0.1666,
+      "step": 672
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.923620367901894e-06,
+      "loss": 0.1732,
+      "step": 673
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.923487070114636e-06,
+      "loss": 0.1691,
+      "step": 674
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.92335377232738e-06,
+      "loss": 0.1647,
+      "step": 675
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.923220474540124e-06,
+      "loss": 0.1806,
+      "step": 676
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.923087176752867e-06,
+      "loss": 0.1578,
+      "step": 677
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.92295387896561e-06,
+      "loss": 0.1852,
+      "step": 678
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.922820581178352e-06,
+      "loss": 0.1855,
+      "step": 679
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.922687283391097e-06,
+      "loss": 0.1756,
+      "step": 680
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.92255398560384e-06,
+      "loss": 0.1588,
+      "step": 681
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.922420687816582e-06,
+      "loss": 0.1507,
+      "step": 682
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.922287390029325e-06,
+      "loss": 0.2001,
+      "step": 683
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.92215409224207e-06,
+      "loss": 0.1605,
+      "step": 684
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.922020794454813e-06,
+      "loss": 0.1402,
+      "step": 685
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.921887496667555e-06,
+      "loss": 0.1552,
+      "step": 686
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.921754198880298e-06,
+      "loss": 0.2262,
+      "step": 687
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.921620901093043e-06,
+      "loss": 0.169,
+      "step": 688
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.921487603305785e-06,
+      "loss": 0.2193,
+      "step": 689
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.921354305518528e-06,
+      "loss": 0.2318,
+      "step": 690
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.921221007731273e-06,
+      "loss": 0.2113,
+      "step": 691
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.921087709944016e-06,
+      "loss": 0.1774,
+      "step": 692
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.920954412156758e-06,
+      "loss": 0.1587,
+      "step": 693
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.920821114369503e-06,
+      "loss": 0.1783,
+      "step": 694
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.920687816582246e-06,
+      "loss": 0.1698,
+      "step": 695
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.920554518794989e-06,
+      "loss": 0.158,
+      "step": 696
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.920421221007731e-06,
+      "loss": 0.1701,
+      "step": 697
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.920287923220476e-06,
+      "loss": 0.1478,
+      "step": 698
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.920154625433219e-06,
+      "loss": 0.1531,
+      "step": 699
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.920021327645962e-06,
+      "loss": 0.1941,
+      "step": 700
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.919888029858706e-06,
+      "loss": 0.194,
+      "step": 701
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.919754732071449e-06,
+      "loss": 0.1539,
+      "step": 702
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.919621434284192e-06,
+      "loss": 0.175,
+      "step": 703
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.919488136496935e-06,
+      "loss": 0.2185,
+      "step": 704
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.919354838709679e-06,
+      "loss": 0.1409,
+      "step": 705
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.919221540922422e-06,
+      "loss": 0.1573,
+      "step": 706
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.919088243135165e-06,
+      "loss": 0.1512,
+      "step": 707
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.91895494534791e-06,
+      "loss": 0.1903,
+      "step": 708
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.918821647560652e-06,
+      "loss": 0.1657,
+      "step": 709
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.918688349773395e-06,
+      "loss": 0.173,
+      "step": 710
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.918555051986138e-06,
+      "loss": 0.1924,
+      "step": 711
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.918421754198882e-06,
+      "loss": 0.198,
+      "step": 712
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.918288456411625e-06,
+      "loss": 0.1581,
+      "step": 713
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.918155158624368e-06,
+      "loss": 0.1938,
+      "step": 714
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.91802186083711e-06,
+      "loss": 0.1719,
+      "step": 715
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.917888563049855e-06,
+      "loss": 0.1976,
+      "step": 716
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.917755265262598e-06,
+      "loss": 0.1244,
+      "step": 717
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.91762196747534e-06,
+      "loss": 0.1582,
+      "step": 718
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.917488669688084e-06,
+      "loss": 0.1243,
+      "step": 719
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.917355371900828e-06,
+      "loss": 0.1607,
+      "step": 720
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.917222074113571e-06,
+      "loss": 0.1691,
+      "step": 721
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.917088776326314e-06,
+      "loss": 0.1509,
+      "step": 722
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.916955478539058e-06,
+      "loss": 0.1372,
+      "step": 723
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.916822180751801e-06,
+      "loss": 0.1876,
+      "step": 724
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.916688882964544e-06,
+      "loss": 0.192,
+      "step": 725
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.916555585177287e-06,
+      "loss": 0.1273,
+      "step": 726
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.916422287390031e-06,
+      "loss": 0.1551,
+      "step": 727
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.916288989602774e-06,
+      "loss": 0.195,
+      "step": 728
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.916155691815517e-06,
+      "loss": 0.1152,
+      "step": 729
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.91602239402826e-06,
+      "loss": 0.1634,
+      "step": 730
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.915889096241004e-06,
+      "loss": 0.1675,
+      "step": 731
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.915755798453747e-06,
+      "loss": 0.1427,
+      "step": 732
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.91562250066649e-06,
+      "loss": 0.1708,
+      "step": 733
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.915489202879233e-06,
+      "loss": 0.1793,
+      "step": 734
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.915355905091977e-06,
+      "loss": 0.1736,
+      "step": 735
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.91522260730472e-06,
+      "loss": 0.168,
+      "step": 736
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.915089309517463e-06,
+      "loss": 0.1774,
+      "step": 737
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.914956011730206e-06,
+      "loss": 0.1505,
+      "step": 738
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.91482271394295e-06,
+      "loss": 0.1394,
+      "step": 739
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.914689416155693e-06,
+      "loss": 0.147,
+      "step": 740
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.914556118368436e-06,
+      "loss": 0.1287,
+      "step": 741
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.914422820581179e-06,
+      "loss": 0.1769,
+      "step": 742
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.914289522793923e-06,
+      "loss": 0.2391,
+      "step": 743
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.914156225006666e-06,
+      "loss": 0.2356,
+      "step": 744
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.914022927219409e-06,
+      "loss": 0.1548,
+      "step": 745
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.913889629432152e-06,
+      "loss": 0.199,
+      "step": 746
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.913756331644896e-06,
+      "loss": 0.155,
+      "step": 747
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.913623033857639e-06,
+      "loss": 0.1913,
+      "step": 748
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.913489736070382e-06,
+      "loss": 0.1378,
+      "step": 749
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.913356438283125e-06,
+      "loss": 0.1195,
+      "step": 750
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.913223140495869e-06,
+      "loss": 0.1736,
+      "step": 751
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.913089842708612e-06,
+      "loss": 0.1635,
+      "step": 752
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.912956544921355e-06,
+      "loss": 0.1774,
+      "step": 753
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.912823247134098e-06,
+      "loss": 0.1656,
+      "step": 754
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.912689949346842e-06,
+      "loss": 0.1455,
+      "step": 755
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.912556651559585e-06,
+      "loss": 0.1396,
+      "step": 756
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.912423353772328e-06,
+      "loss": 0.2479,
+      "step": 757
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.91229005598507e-06,
+      "loss": 0.1565,
+      "step": 758
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.912156758197815e-06,
+      "loss": 0.1911,
+      "step": 759
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.912023460410558e-06,
+      "loss": 0.2161,
+      "step": 760
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.9118901626233e-06,
+      "loss": 0.1451,
+      "step": 761
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.911756864836043e-06,
+      "loss": 0.2208,
+      "step": 762
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.911623567048788e-06,
+      "loss": 0.1458,
+      "step": 763
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.91149026926153e-06,
+      "loss": 0.202,
+      "step": 764
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.911356971474274e-06,
+      "loss": 0.1551,
+      "step": 765
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.911223673687016e-06,
+      "loss": 0.176,
+      "step": 766
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.911090375899761e-06,
+      "loss": 0.1734,
+      "step": 767
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.910957078112504e-06,
+      "loss": 0.1802,
+      "step": 768
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.910823780325247e-06,
+      "loss": 0.1511,
+      "step": 769
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.91069048253799e-06,
+      "loss": 0.1885,
+      "step": 770
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.910557184750734e-06,
+      "loss": 0.1912,
+      "step": 771
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.910423886963477e-06,
+      "loss": 0.1953,
+      "step": 772
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.91029058917622e-06,
+      "loss": 0.1476,
+      "step": 773
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.910157291388962e-06,
+      "loss": 0.1713,
+      "step": 774
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.910023993601707e-06,
+      "loss": 0.1687,
+      "step": 775
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.90989069581445e-06,
+      "loss": 0.1525,
+      "step": 776
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.909757398027193e-06,
+      "loss": 0.1545,
+      "step": 777
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.909624100239937e-06,
+      "loss": 0.1789,
+      "step": 778
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.90949080245268e-06,
+      "loss": 0.1804,
+      "step": 779
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.909357504665423e-06,
+      "loss": 0.1265,
+      "step": 780
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.909224206878165e-06,
+      "loss": 0.1676,
+      "step": 781
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.90909090909091e-06,
+      "loss": 0.2089,
+      "step": 782
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.908957611303653e-06,
+      "loss": 0.196,
+      "step": 783
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.908824313516396e-06,
+      "loss": 0.1953,
+      "step": 784
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.90869101572914e-06,
+      "loss": 0.2139,
+      "step": 785
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.908557717941883e-06,
+      "loss": 0.1778,
+      "step": 786
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.908424420154626e-06,
+      "loss": 0.1574,
+      "step": 787
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.908291122367369e-06,
+      "loss": 0.1877,
+      "step": 788
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.908157824580113e-06,
+      "loss": 0.1747,
+      "step": 789
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.908024526792856e-06,
+      "loss": 0.1891,
+      "step": 790
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.907891229005599e-06,
+      "loss": 0.1897,
+      "step": 791
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.907757931218343e-06,
+      "loss": 0.1134,
+      "step": 792
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.907624633431086e-06,
+      "loss": 0.15,
+      "step": 793
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.907491335643829e-06,
+      "loss": 0.1782,
+      "step": 794
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.907358037856572e-06,
+      "loss": 0.1685,
+      "step": 795
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.907224740069316e-06,
+      "loss": 0.1636,
+      "step": 796
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.907091442282059e-06,
+      "loss": 0.1961,
+      "step": 797
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.906958144494802e-06,
+      "loss": 0.1867,
+      "step": 798
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.906824846707546e-06,
+      "loss": 0.1404,
+      "step": 799
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.90669154892029e-06,
+      "loss": 0.1646,
+      "step": 800
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.906558251133032e-06,
+      "loss": 0.1574,
+      "step": 801
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.906424953345775e-06,
+      "loss": 0.1505,
+      "step": 802
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.90629165555852e-06,
+      "loss": 0.1298,
+      "step": 803
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.906158357771262e-06,
+      "loss": 0.1465,
+      "step": 804
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.906025059984005e-06,
+      "loss": 0.1591,
+      "step": 805
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.90589176219675e-06,
+      "loss": 0.1406,
+      "step": 806
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.905758464409492e-06,
+      "loss": 0.2217,
+      "step": 807
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.905625166622235e-06,
+      "loss": 0.1988,
+      "step": 808
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.905491868834978e-06,
+      "loss": 0.152,
+      "step": 809
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.905358571047722e-06,
+      "loss": 0.1957,
+      "step": 810
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.905225273260465e-06,
+      "loss": 0.1404,
+      "step": 811
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.905091975473208e-06,
+      "loss": 0.137,
+      "step": 812
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.904958677685951e-06,
+      "loss": 0.1345,
+      "step": 813
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.904825379898695e-06,
+      "loss": 0.1551,
+      "step": 814
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.904692082111438e-06,
+      "loss": 0.2091,
+      "step": 815
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.904558784324181e-06,
+      "loss": 0.1576,
+      "step": 816
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.904425486536924e-06,
+      "loss": 0.137,
+      "step": 817
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.904292188749668e-06,
+      "loss": 0.1705,
+      "step": 818
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.904158890962411e-06,
+      "loss": 0.1848,
+      "step": 819
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.904025593175154e-06,
+      "loss": 0.1729,
+      "step": 820
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.903892295387897e-06,
+      "loss": 0.1907,
+      "step": 821
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.903758997600641e-06,
+      "loss": 0.1956,
+      "step": 822
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.903625699813384e-06,
+      "loss": 0.1829,
+      "step": 823
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.903492402026127e-06,
+      "loss": 0.1181,
+      "step": 824
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.90335910423887e-06,
+      "loss": 0.1403,
+      "step": 825
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.903225806451614e-06,
+      "loss": 0.165,
+      "step": 826
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.903092508664357e-06,
+      "loss": 0.1747,
+      "step": 827
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.9029592108771e-06,
+      "loss": 0.1495,
+      "step": 828
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.902825913089843e-06,
+      "loss": 0.1512,
+      "step": 829
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.902692615302587e-06,
+      "loss": 0.1567,
+      "step": 830
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.90255931751533e-06,
+      "loss": 0.1674,
+      "step": 831
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.902426019728073e-06,
+      "loss": 0.1558,
+      "step": 832
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.902292721940817e-06,
+      "loss": 0.143,
+      "step": 833
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.90215942415356e-06,
+      "loss": 0.1314,
+      "step": 834
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.902026126366303e-06,
+      "loss": 0.1732,
+      "step": 835
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.901892828579046e-06,
+      "loss": 0.1208,
+      "step": 836
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.90175953079179e-06,
+      "loss": 0.1235,
+      "step": 837
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.901626233004533e-06,
+      "loss": 0.1178,
+      "step": 838
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.901492935217276e-06,
+      "loss": 0.1831,
+      "step": 839
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.901359637430019e-06,
+      "loss": 0.1846,
+      "step": 840
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.901226339642763e-06,
+      "loss": 0.1738,
+      "step": 841
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.901093041855506e-06,
+      "loss": 0.1703,
+      "step": 842
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.900959744068249e-06,
+      "loss": 0.1429,
+      "step": 843
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.900826446280992e-06,
+      "loss": 0.1382,
+      "step": 844
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.900693148493736e-06,
+      "loss": 0.154,
+      "step": 845
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.90055985070648e-06,
+      "loss": 0.1789,
+      "step": 846
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.900426552919222e-06,
+      "loss": 0.1653,
+      "step": 847
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.900293255131965e-06,
+      "loss": 0.1556,
+      "step": 848
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.90015995734471e-06,
+      "loss": 0.1255,
+      "step": 849
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.900026659557452e-06,
+      "loss": 0.1751,
+      "step": 850
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.899893361770195e-06,
+      "loss": 0.1888,
+      "step": 851
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.899760063982938e-06,
+      "loss": 0.1519,
+      "step": 852
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.899626766195682e-06,
+      "loss": 0.1854,
+      "step": 853
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.899493468408425e-06,
+      "loss": 0.1454,
+      "step": 854
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.899360170621168e-06,
+      "loss": 0.1543,
+      "step": 855
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.89922687283391e-06,
+      "loss": 0.1302,
+      "step": 856
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.899093575046655e-06,
+      "loss": 0.1612,
+      "step": 857
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.898960277259398e-06,
+      "loss": 0.1413,
+      "step": 858
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.898826979472141e-06,
+      "loss": 0.1887,
+      "step": 859
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.898693681684884e-06,
+      "loss": 0.1204,
+      "step": 860
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.898560383897628e-06,
+      "loss": 0.2046,
+      "step": 861
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.898427086110371e-06,
+      "loss": 0.1633,
+      "step": 862
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.898293788323114e-06,
+      "loss": 0.1758,
+      "step": 863
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.898160490535857e-06,
+      "loss": 0.1626,
+      "step": 864
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.898027192748601e-06,
+      "loss": 0.1279,
+      "step": 865
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.897893894961344e-06,
+      "loss": 0.1581,
+      "step": 866
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.897760597174087e-06,
+      "loss": 0.177,
+      "step": 867
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.89762729938683e-06,
+      "loss": 0.157,
+      "step": 868
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.897494001599574e-06,
+      "loss": 0.1084,
+      "step": 869
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.897360703812317e-06,
+      "loss": 0.1831,
+      "step": 870
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.89722740602506e-06,
+      "loss": 0.1663,
+      "step": 871
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.897094108237804e-06,
+      "loss": 0.176,
+      "step": 872
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.896960810450547e-06,
+      "loss": 0.1734,
+      "step": 873
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.89682751266329e-06,
+      "loss": 0.1528,
+      "step": 874
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.896694214876033e-06,
+      "loss": 0.1932,
+      "step": 875
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.896560917088777e-06,
+      "loss": 0.1835,
+      "step": 876
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.89642761930152e-06,
+      "loss": 0.1523,
+      "step": 877
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.896294321514263e-06,
+      "loss": 0.1522,
+      "step": 878
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.896161023727007e-06,
+      "loss": 0.1885,
+      "step": 879
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.89602772593975e-06,
+      "loss": 0.1921,
+      "step": 880
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.895894428152493e-06,
+      "loss": 0.2162,
+      "step": 881
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.895761130365236e-06,
+      "loss": 0.1617,
+      "step": 882
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.89562783257798e-06,
+      "loss": 0.1266,
+      "step": 883
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.895494534790723e-06,
+      "loss": 0.1402,
+      "step": 884
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.895361237003466e-06,
+      "loss": 0.1738,
+      "step": 885
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.89522793921621e-06,
+      "loss": 0.1417,
+      "step": 886
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.895094641428953e-06,
+      "loss": 0.2088,
+      "step": 887
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.894961343641696e-06,
+      "loss": 0.166,
+      "step": 888
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.894828045854439e-06,
+      "loss": 0.165,
+      "step": 889
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.894694748067184e-06,
+      "loss": 0.1643,
+      "step": 890
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.894561450279926e-06,
+      "loss": 0.1648,
+      "step": 891
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.89442815249267e-06,
+      "loss": 0.154,
+      "step": 892
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.894294854705414e-06,
+      "loss": 0.1594,
+      "step": 893
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.894161556918157e-06,
+      "loss": 0.1798,
+      "step": 894
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.8940282591309e-06,
+      "loss": 0.1141,
+      "step": 895
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.893894961343642e-06,
+      "loss": 0.1377,
+      "step": 896
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.893761663556387e-06,
+      "loss": 0.1692,
+      "step": 897
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.89362836576913e-06,
+      "loss": 0.1552,
+      "step": 898
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.893495067981872e-06,
+      "loss": 0.167,
+      "step": 899
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.893361770194617e-06,
+      "loss": 0.146,
+      "step": 900
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.89322847240736e-06,
+      "loss": 0.1355,
+      "step": 901
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.893095174620103e-06,
+      "loss": 0.1307,
+      "step": 902
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.892961876832845e-06,
+      "loss": 0.1999,
+      "step": 903
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.89282857904559e-06,
+      "loss": 0.1594,
+      "step": 904
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.892695281258333e-06,
+      "loss": 0.1921,
+      "step": 905
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.892561983471075e-06,
+      "loss": 0.1769,
+      "step": 906
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.892428685683818e-06,
+      "loss": 0.1411,
+      "step": 907
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.892295387896563e-06,
+      "loss": 0.1889,
+      "step": 908
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.892162090109306e-06,
+      "loss": 0.1375,
+      "step": 909
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.892028792322048e-06,
+      "loss": 0.1348,
+      "step": 910
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.891895494534791e-06,
+      "loss": 0.1877,
+      "step": 911
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.891762196747536e-06,
+      "loss": 0.1247,
+      "step": 912
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.891628898960279e-06,
+      "loss": 0.126,
+      "step": 913
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.891495601173021e-06,
+      "loss": 0.1373,
+      "step": 914
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.891362303385764e-06,
+      "loss": 0.1543,
+      "step": 915
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.891229005598509e-06,
+      "loss": 0.1688,
+      "step": 916
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.891095707811252e-06,
+      "loss": 0.1385,
+      "step": 917
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.890962410023994e-06,
+      "loss": 0.1391,
+      "step": 918
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.890829112236737e-06,
+      "loss": 0.1521,
+      "step": 919
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.890695814449482e-06,
+      "loss": 0.1645,
+      "step": 920
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.890562516662225e-06,
+      "loss": 0.1725,
+      "step": 921
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.890429218874967e-06,
+      "loss": 0.1651,
+      "step": 922
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.89029592108771e-06,
+      "loss": 0.1621,
+      "step": 923
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.890162623300455e-06,
+      "loss": 0.1441,
+      "step": 924
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.890029325513198e-06,
+      "loss": 0.1976,
+      "step": 925
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.88989602772594e-06,
+      "loss": 0.1737,
+      "step": 926
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.889762729938683e-06,
+      "loss": 0.139,
+      "step": 927
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.889629432151428e-06,
+      "loss": 0.1533,
+      "step": 928
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.88949613436417e-06,
+      "loss": 0.1576,
+      "step": 929
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.889362836576913e-06,
+      "loss": 0.1375,
+      "step": 930
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.889229538789656e-06,
+      "loss": 0.2009,
+      "step": 931
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.8890962410024e-06,
+      "loss": 0.1409,
+      "step": 932
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.888962943215143e-06,
+      "loss": 0.1687,
+      "step": 933
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.888829645427886e-06,
+      "loss": 0.2027,
+      "step": 934
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.888696347640629e-06,
+      "loss": 0.1772,
+      "step": 935
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.888563049853374e-06,
+      "loss": 0.1393,
+      "step": 936
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.888429752066116e-06,
+      "loss": 0.1818,
+      "step": 937
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.88829645427886e-06,
+      "loss": 0.1454,
+      "step": 938
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.888163156491602e-06,
+      "loss": 0.1536,
+      "step": 939
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.888029858704347e-06,
+      "loss": 0.139,
+      "step": 940
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.88789656091709e-06,
+      "loss": 0.1459,
+      "step": 941
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.887763263129832e-06,
+      "loss": 0.1655,
+      "step": 942
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.887629965342575e-06,
+      "loss": 0.1524,
+      "step": 943
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.88749666755532e-06,
+      "loss": 0.1415,
+      "step": 944
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.887363369768062e-06,
+      "loss": 0.1269,
+      "step": 945
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.887230071980805e-06,
+      "loss": 0.1898,
+      "step": 946
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.88709677419355e-06,
+      "loss": 0.1836,
+      "step": 947
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.886963476406293e-06,
+      "loss": 0.137,
+      "step": 948
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.886830178619035e-06,
+      "loss": 0.1741,
+      "step": 949
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.886696880831778e-06,
+      "loss": 0.1577,
+      "step": 950
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.886563583044523e-06,
+      "loss": 0.1221,
+      "step": 951
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.886430285257265e-06,
+      "loss": 0.1395,
+      "step": 952
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.886296987470008e-06,
+      "loss": 0.1467,
+      "step": 953
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.886163689682751e-06,
+      "loss": 0.106,
+      "step": 954
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.886030391895496e-06,
+      "loss": 0.1216,
+      "step": 955
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.885897094108238e-06,
+      "loss": 0.1313,
+      "step": 956
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.885763796320981e-06,
+      "loss": 0.1573,
+      "step": 957
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.885630498533724e-06,
+      "loss": 0.1614,
+      "step": 958
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.885497200746469e-06,
+      "loss": 0.1203,
+      "step": 959
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.885363902959211e-06,
+      "loss": 0.1798,
+      "step": 960
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.885230605171954e-06,
+      "loss": 0.1304,
+      "step": 961
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.885097307384697e-06,
+      "loss": 0.1326,
+      "step": 962
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.884964009597442e-06,
+      "loss": 0.1463,
+      "step": 963
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.884830711810184e-06,
+      "loss": 0.2028,
+      "step": 964
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.884697414022927e-06,
+      "loss": 0.1749,
+      "step": 965
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.884564116235672e-06,
+      "loss": 0.1686,
+      "step": 966
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.884430818448415e-06,
+      "loss": 0.1643,
+      "step": 967
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.884297520661157e-06,
+      "loss": 0.1534,
+      "step": 968
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.8841642228739e-06,
+      "loss": 0.165,
+      "step": 969
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.884030925086645e-06,
+      "loss": 0.1495,
+      "step": 970
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.883897627299388e-06,
+      "loss": 0.1731,
+      "step": 971
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.88376432951213e-06,
+      "loss": 0.146,
+      "step": 972
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.883631031724875e-06,
+      "loss": 0.2028,
+      "step": 973
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.883497733937618e-06,
+      "loss": 0.1138,
+      "step": 974
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.88336443615036e-06,
+      "loss": 0.1541,
+      "step": 975
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.883231138363103e-06,
+      "loss": 0.14,
+      "step": 976
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.883097840575848e-06,
+      "loss": 0.149,
+      "step": 977
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.88296454278859e-06,
+      "loss": 0.1504,
+      "step": 978
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.882831245001333e-06,
+      "loss": 0.1369,
+      "step": 979
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.882697947214078e-06,
+      "loss": 0.1189,
+      "step": 980
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.88256464942682e-06,
+      "loss": 0.156,
+      "step": 981
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.882431351639564e-06,
+      "loss": 0.1359,
+      "step": 982
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.882298053852306e-06,
+      "loss": 0.1437,
+      "step": 983
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.882164756065051e-06,
+      "loss": 0.1457,
+      "step": 984
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.882031458277794e-06,
+      "loss": 0.1843,
+      "step": 985
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.881898160490537e-06,
+      "loss": 0.1888,
+      "step": 986
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.881764862703281e-06,
+      "loss": 0.1687,
+      "step": 987
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.881631564916024e-06,
+      "loss": 0.1469,
+      "step": 988
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.881498267128767e-06,
+      "loss": 0.1636,
+      "step": 989
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.88136496934151e-06,
+      "loss": 0.1467,
+      "step": 990
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.881231671554254e-06,
+      "loss": 0.1439,
+      "step": 991
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.881098373766997e-06,
+      "loss": 0.1905,
+      "step": 992
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.88096507597974e-06,
+      "loss": 0.1454,
+      "step": 993
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.880831778192483e-06,
+      "loss": 0.1379,
+      "step": 994
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.880698480405227e-06,
+      "loss": 0.1412,
+      "step": 995
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.88056518261797e-06,
+      "loss": 0.121,
+      "step": 996
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.880431884830713e-06,
+      "loss": 0.1289,
+      "step": 997
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.880298587043455e-06,
+      "loss": 0.1393,
+      "step": 998
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.8801652892562e-06,
+      "loss": 0.2252,
+      "step": 999
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.880031991468943e-06,
+      "loss": 0.1656,
+      "step": 1000
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.879898693681686e-06,
+      "loss": 0.1667,
+      "step": 1001
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.87976539589443e-06,
+      "loss": 0.1027,
+      "step": 1002
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.879632098107173e-06,
+      "loss": 0.1558,
+      "step": 1003
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.879498800319916e-06,
+      "loss": 0.1078,
+      "step": 1004
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.879365502532659e-06,
+      "loss": 0.1656,
+      "step": 1005
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.879232204745403e-06,
+      "loss": 0.1897,
+      "step": 1006
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.879098906958146e-06,
+      "loss": 0.1746,
+      "step": 1007
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.878965609170889e-06,
+      "loss": 0.122,
+      "step": 1008
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.878832311383632e-06,
+      "loss": 0.1619,
+      "step": 1009
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.878699013596376e-06,
+      "loss": 0.1567,
+      "step": 1010
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.878565715809119e-06,
+      "loss": 0.17,
+      "step": 1011
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.878432418021862e-06,
+      "loss": 0.1647,
+      "step": 1012
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.878299120234605e-06,
+      "loss": 0.1527,
+      "step": 1013
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.878165822447349e-06,
+      "loss": 0.1189,
+      "step": 1014
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.878032524660092e-06,
+      "loss": 0.1575,
+      "step": 1015
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.877899226872835e-06,
+      "loss": 0.1438,
+      "step": 1016
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.877765929085578e-06,
+      "loss": 0.1432,
+      "step": 1017
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.877632631298322e-06,
+      "loss": 0.1305,
+      "step": 1018
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.877499333511065e-06,
+      "loss": 0.146,
+      "step": 1019
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.877366035723808e-06,
+      "loss": 0.1498,
+      "step": 1020
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.87723273793655e-06,
+      "loss": 0.2041,
+      "step": 1021
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.877099440149295e-06,
+      "loss": 0.1359,
+      "step": 1022
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.876966142362038e-06,
+      "loss": 0.1605,
+      "step": 1023
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.87683284457478e-06,
+      "loss": 0.1448,
+      "step": 1024
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.876699546787523e-06,
+      "loss": 0.1058,
+      "step": 1025
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.876566249000268e-06,
+      "loss": 0.1357,
+      "step": 1026
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.87643295121301e-06,
+      "loss": 0.122,
+      "step": 1027
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.876299653425754e-06,
+      "loss": 0.1558,
+      "step": 1028
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.876166355638496e-06,
+      "loss": 0.1536,
+      "step": 1029
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.876033057851241e-06,
+      "loss": 0.1356,
+      "step": 1030
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.875899760063984e-06,
+      "loss": 0.1609,
+      "step": 1031
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.875766462276727e-06,
+      "loss": 0.1555,
+      "step": 1032
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.87563316448947e-06,
+      "loss": 0.1918,
+      "step": 1033
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.875499866702214e-06,
+      "loss": 0.1461,
+      "step": 1034
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.875366568914957e-06,
+      "loss": 0.1577,
+      "step": 1035
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.8752332711277e-06,
+      "loss": 0.1437,
+      "step": 1036
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.875099973340442e-06,
+      "loss": 0.1584,
+      "step": 1037
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.874966675553187e-06,
+      "loss": 0.1548,
+      "step": 1038
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.87483337776593e-06,
+      "loss": 0.1649,
+      "step": 1039
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.874700079978673e-06,
+      "loss": 0.1379,
+      "step": 1040
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.874566782191415e-06,
+      "loss": 0.1658,
+      "step": 1041
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.87443348440416e-06,
+      "loss": 0.1699,
+      "step": 1042
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.874300186616903e-06,
+      "loss": 0.1507,
+      "step": 1043
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.874166888829645e-06,
+      "loss": 0.1701,
+      "step": 1044
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.874033591042388e-06,
+      "loss": 0.1712,
+      "step": 1045
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.873900293255133e-06,
+      "loss": 0.1389,
+      "step": 1046
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.873766995467876e-06,
+      "loss": 0.1768,
+      "step": 1047
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.873633697680618e-06,
+      "loss": 0.1199,
+      "step": 1048
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.873500399893361e-06,
+      "loss": 0.207,
+      "step": 1049
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.873367102106106e-06,
+      "loss": 0.1657,
+      "step": 1050
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.873233804318849e-06,
+      "loss": 0.1794,
+      "step": 1051
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.873100506531591e-06,
+      "loss": 0.2125,
+      "step": 1052
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.872967208744336e-06,
+      "loss": 0.1473,
+      "step": 1053
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.872833910957079e-06,
+      "loss": 0.1697,
+      "step": 1054
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.872700613169822e-06,
+      "loss": 0.1718,
+      "step": 1055
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.872567315382564e-06,
+      "loss": 0.1392,
+      "step": 1056
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.872434017595309e-06,
+      "loss": 0.1626,
+      "step": 1057
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.872300719808052e-06,
+      "loss": 0.1343,
+      "step": 1058
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.872167422020795e-06,
+      "loss": 0.1896,
+      "step": 1059
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.872034124233539e-06,
+      "loss": 0.1608,
+      "step": 1060
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.871900826446282e-06,
+      "loss": 0.1361,
+      "step": 1061
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.871767528659025e-06,
+      "loss": 0.1741,
+      "step": 1062
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.871634230871768e-06,
+      "loss": 0.157,
+      "step": 1063
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.871500933084512e-06,
+      "loss": 0.1663,
+      "step": 1064
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.871367635297255e-06,
+      "loss": 0.1395,
+      "step": 1065
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.871234337509998e-06,
+      "loss": 0.1365,
+      "step": 1066
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.871101039722742e-06,
+      "loss": 0.1467,
+      "step": 1067
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.870967741935485e-06,
+      "loss": 0.1983,
+      "step": 1068
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.870834444148228e-06,
+      "loss": 0.1699,
+      "step": 1069
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.87070114636097e-06,
+      "loss": 0.1379,
+      "step": 1070
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.870567848573715e-06,
+      "loss": 0.1887,
+      "step": 1071
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.870434550786458e-06,
+      "loss": 0.1458,
+      "step": 1072
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.8703012529992e-06,
+      "loss": 0.1449,
+      "step": 1073
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.870167955211945e-06,
+      "loss": 0.1263,
+      "step": 1074
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.870034657424688e-06,
+      "loss": 0.1219,
+      "step": 1075
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.869901359637431e-06,
+      "loss": 0.1675,
+      "step": 1076
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.869768061850174e-06,
+      "loss": 0.1802,
+      "step": 1077
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.869634764062918e-06,
+      "loss": 0.1632,
+      "step": 1078
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.869501466275661e-06,
+      "loss": 0.1478,
+      "step": 1079
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.869368168488404e-06,
+      "loss": 0.2206,
+      "step": 1080
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.869234870701148e-06,
+      "loss": 0.1582,
+      "step": 1081
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.869101572913891e-06,
+      "loss": 0.1417,
+      "step": 1082
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.868968275126634e-06,
+      "loss": 0.1686,
+      "step": 1083
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.868834977339377e-06,
+      "loss": 0.1417,
+      "step": 1084
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.868701679552121e-06,
+      "loss": 0.1921,
+      "step": 1085
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.868568381764864e-06,
+      "loss": 0.1885,
+      "step": 1086
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.868435083977607e-06,
+      "loss": 0.1454,
+      "step": 1087
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.86830178619035e-06,
+      "loss": 0.1564,
+      "step": 1088
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.868168488403094e-06,
+      "loss": 0.1156,
+      "step": 1089
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.868035190615837e-06,
+      "loss": 0.1387,
+      "step": 1090
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.86790189282858e-06,
+      "loss": 0.1473,
+      "step": 1091
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.867768595041323e-06,
+      "loss": 0.1313,
+      "step": 1092
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.867635297254067e-06,
+      "loss": 0.1569,
+      "step": 1093
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.86750199946681e-06,
+      "loss": 0.1378,
+      "step": 1094
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.867368701679553e-06,
+      "loss": 0.1263,
+      "step": 1095
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.867235403892296e-06,
+      "loss": 0.1853,
+      "step": 1096
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.86710210610504e-06,
+      "loss": 0.1157,
+      "step": 1097
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.866968808317783e-06,
+      "loss": 0.1615,
+      "step": 1098
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.866835510530526e-06,
+      "loss": 0.1504,
+      "step": 1099
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.866702212743269e-06,
+      "loss": 0.2074,
+      "step": 1100
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.866568914956013e-06,
+      "loss": 0.1364,
+      "step": 1101
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.866435617168756e-06,
+      "loss": 0.1417,
+      "step": 1102
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.866302319381499e-06,
+      "loss": 0.1986,
+      "step": 1103
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.866169021594242e-06,
+      "loss": 0.1466,
+      "step": 1104
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.866035723806986e-06,
+      "loss": 0.0855,
+      "step": 1105
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.865902426019729e-06,
+      "loss": 0.1666,
+      "step": 1106
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.865769128232472e-06,
+      "loss": 0.1177,
+      "step": 1107
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.865635830445215e-06,
+      "loss": 0.1829,
+      "step": 1108
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.86550253265796e-06,
+      "loss": 0.1063,
+      "step": 1109
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.865369234870702e-06,
+      "loss": 0.1191,
+      "step": 1110
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.865235937083445e-06,
+      "loss": 0.1326,
+      "step": 1111
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.865102639296188e-06,
+      "loss": 0.1228,
+      "step": 1112
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.864969341508932e-06,
+      "loss": 0.1285,
+      "step": 1113
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.864836043721675e-06,
+      "loss": 0.1195,
+      "step": 1114
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.864702745934418e-06,
+      "loss": 0.222,
+      "step": 1115
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.864569448147162e-06,
+      "loss": 0.1306,
+      "step": 1116
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.864436150359905e-06,
+      "loss": 0.1237,
+      "step": 1117
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.864302852572648e-06,
+      "loss": 0.1729,
+      "step": 1118
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.86416955478539e-06,
+      "loss": 0.1586,
+      "step": 1119
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.864036256998135e-06,
+      "loss": 0.1675,
+      "step": 1120
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.863902959210878e-06,
+      "loss": 0.1437,
+      "step": 1121
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.863769661423621e-06,
+      "loss": 0.143,
+      "step": 1122
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.863636363636364e-06,
+      "loss": 0.1511,
+      "step": 1123
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.863503065849108e-06,
+      "loss": 0.1275,
+      "step": 1124
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.863369768061851e-06,
+      "loss": 0.1726,
+      "step": 1125
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.863236470274594e-06,
+      "loss": 0.1357,
+      "step": 1126
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.863103172487337e-06,
+      "loss": 0.1299,
+      "step": 1127
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.862969874700081e-06,
+      "loss": 0.1288,
+      "step": 1128
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.862836576912824e-06,
+      "loss": 0.1784,
+      "step": 1129
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.862703279125567e-06,
+      "loss": 0.1633,
+      "step": 1130
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.86256998133831e-06,
+      "loss": 0.1857,
+      "step": 1131
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.862436683551054e-06,
+      "loss": 0.1715,
+      "step": 1132
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.862303385763797e-06,
+      "loss": 0.1449,
+      "step": 1133
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.86217008797654e-06,
+      "loss": 0.1308,
+      "step": 1134
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.862036790189283e-06,
+      "loss": 0.1578,
+      "step": 1135
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.861903492402027e-06,
+      "loss": 0.1262,
+      "step": 1136
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.86177019461477e-06,
+      "loss": 0.1631,
+      "step": 1137
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.861636896827513e-06,
+      "loss": 0.1701,
+      "step": 1138
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.861503599040256e-06,
+      "loss": 0.1764,
+      "step": 1139
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.861370301253e-06,
+      "loss": 0.1564,
+      "step": 1140
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.861237003465743e-06,
+      "loss": 0.1483,
+      "step": 1141
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.861103705678486e-06,
+      "loss": 0.1488,
+      "step": 1142
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.860970407891229e-06,
+      "loss": 0.1641,
+      "step": 1143
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.860837110103973e-06,
+      "loss": 0.2019,
+      "step": 1144
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.860703812316716e-06,
+      "loss": 0.2056,
+      "step": 1145
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.860570514529459e-06,
+      "loss": 0.1638,
+      "step": 1146
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.860437216742203e-06,
+      "loss": 0.1602,
+      "step": 1147
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.860303918954946e-06,
+      "loss": 0.1615,
+      "step": 1148
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.860170621167689e-06,
+      "loss": 0.1504,
+      "step": 1149
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.860037323380432e-06,
+      "loss": 0.1617,
+      "step": 1150
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.859904025593176e-06,
+      "loss": 0.205,
+      "step": 1151
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.859770727805919e-06,
+      "loss": 0.1519,
+      "step": 1152
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.859637430018662e-06,
+      "loss": 0.154,
+      "step": 1153
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.859504132231406e-06,
+      "loss": 0.1129,
+      "step": 1154
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.85937083444415e-06,
+      "loss": 0.1418,
+      "step": 1155
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.859237536656892e-06,
+      "loss": 0.1684,
+      "step": 1156
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.859104238869635e-06,
+      "loss": 0.1165,
+      "step": 1157
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.85897094108238e-06,
+      "loss": 0.1074,
+      "step": 1158
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.858837643295122e-06,
+      "loss": 0.1393,
+      "step": 1159
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.858704345507865e-06,
+      "loss": 0.1383,
+      "step": 1160
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.85857104772061e-06,
+      "loss": 0.1408,
+      "step": 1161
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.858437749933352e-06,
+      "loss": 0.1456,
+      "step": 1162
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.858304452146095e-06,
+      "loss": 0.1634,
+      "step": 1163
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.858171154358838e-06,
+      "loss": 0.1793,
+      "step": 1164
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.858037856571582e-06,
+      "loss": 0.1347,
+      "step": 1165
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.857904558784325e-06,
+      "loss": 0.1362,
+      "step": 1166
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.857771260997068e-06,
+      "loss": 0.1466,
+      "step": 1167
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.857637963209813e-06,
+      "loss": 0.1216,
+      "step": 1168
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.857504665422555e-06,
+      "loss": 0.112,
+      "step": 1169
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.857371367635298e-06,
+      "loss": 0.2115,
+      "step": 1170
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.857238069848041e-06,
+      "loss": 0.1558,
+      "step": 1171
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.857104772060786e-06,
+      "loss": 0.144,
+      "step": 1172
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.856971474273528e-06,
+      "loss": 0.1741,
+      "step": 1173
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.856838176486271e-06,
+      "loss": 0.146,
+      "step": 1174
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.856704878699016e-06,
+      "loss": 0.1828,
+      "step": 1175
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.856571580911759e-06,
+      "loss": 0.1692,
+      "step": 1176
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.856438283124501e-06,
+      "loss": 0.1688,
+      "step": 1177
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.856304985337244e-06,
+      "loss": 0.1398,
+      "step": 1178
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.856171687549989e-06,
+      "loss": 0.1627,
+      "step": 1179
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.856038389762732e-06,
+      "loss": 0.1726,
+      "step": 1180
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.855905091975474e-06,
+      "loss": 0.1704,
+      "step": 1181
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.855771794188217e-06,
+      "loss": 0.1605,
+      "step": 1182
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.855638496400962e-06,
+      "loss": 0.1453,
+      "step": 1183
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.855505198613705e-06,
+      "loss": 0.2029,
+      "step": 1184
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.855371900826447e-06,
+      "loss": 0.1377,
+      "step": 1185
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.85523860303919e-06,
+      "loss": 0.1554,
+      "step": 1186
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.855105305251935e-06,
+      "loss": 0.1576,
+      "step": 1187
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.854972007464677e-06,
+      "loss": 0.1807,
+      "step": 1188
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.85483870967742e-06,
+      "loss": 0.1733,
+      "step": 1189
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.854705411890163e-06,
+      "loss": 0.2,
+      "step": 1190
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.854572114102908e-06,
+      "loss": 0.1706,
+      "step": 1191
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.85443881631565e-06,
+      "loss": 0.1726,
+      "step": 1192
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.854305518528393e-06,
+      "loss": 0.1466,
+      "step": 1193
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.854172220741136e-06,
+      "loss": 0.1799,
+      "step": 1194
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.85403892295388e-06,
+      "loss": 0.1938,
+      "step": 1195
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.853905625166623e-06,
+      "loss": 0.1515,
+      "step": 1196
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.853772327379366e-06,
+      "loss": 0.1618,
+      "step": 1197
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.853639029592109e-06,
+      "loss": 0.1426,
+      "step": 1198
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.853505731804854e-06,
+      "loss": 0.1661,
+      "step": 1199
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.853372434017596e-06,
+      "loss": 0.1523,
+      "step": 1200
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.85323913623034e-06,
+      "loss": 0.1323,
+      "step": 1201
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.853105838443082e-06,
+      "loss": 0.1203,
+      "step": 1202
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.852972540655827e-06,
+      "loss": 0.1625,
+      "step": 1203
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.85283924286857e-06,
+      "loss": 0.123,
+      "step": 1204
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.852705945081312e-06,
+      "loss": 0.1426,
+      "step": 1205
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.852572647294055e-06,
+      "loss": 0.1143,
+      "step": 1206
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.8524393495068e-06,
+      "loss": 0.13,
+      "step": 1207
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.852306051719542e-06,
+      "loss": 0.1552,
+      "step": 1208
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.852172753932285e-06,
+      "loss": 0.1566,
+      "step": 1209
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.852039456145028e-06,
+      "loss": 0.0947,
+      "step": 1210
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.851906158357772e-06,
+      "loss": 0.1307,
+      "step": 1211
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.851772860570515e-06,
+      "loss": 0.1406,
+      "step": 1212
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.851639562783258e-06,
+      "loss": 0.1437,
+      "step": 1213
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.851506264996001e-06,
+      "loss": 0.1866,
+      "step": 1214
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.851372967208745e-06,
+      "loss": 0.1453,
+      "step": 1215
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.851239669421488e-06,
+      "loss": 0.1545,
+      "step": 1216
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.851106371634231e-06,
+      "loss": 0.1434,
+      "step": 1217
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.850973073846974e-06,
+      "loss": 0.176,
+      "step": 1218
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.850839776059718e-06,
+      "loss": 0.1788,
+      "step": 1219
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.850706478272461e-06,
+      "loss": 0.1622,
+      "step": 1220
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.850573180485204e-06,
+      "loss": 0.1235,
+      "step": 1221
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.850439882697947e-06,
+      "loss": 0.1851,
+      "step": 1222
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.850306584910691e-06,
+      "loss": 0.1311,
+      "step": 1223
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.850173287123434e-06,
+      "loss": 0.1262,
+      "step": 1224
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.850039989336177e-06,
+      "loss": 0.1642,
+      "step": 1225
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.84990669154892e-06,
+      "loss": 0.148,
+      "step": 1226
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.849773393761664e-06,
+      "loss": 0.124,
+      "step": 1227
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.849640095974407e-06,
+      "loss": 0.1405,
+      "step": 1228
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.84950679818715e-06,
+      "loss": 0.1409,
+      "step": 1229
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.849373500399895e-06,
+      "loss": 0.1779,
+      "step": 1230
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.849240202612637e-06,
+      "loss": 0.1166,
+      "step": 1231
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.84910690482538e-06,
+      "loss": 0.1556,
+      "step": 1232
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.848973607038123e-06,
+      "loss": 0.1369,
+      "step": 1233
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.848840309250867e-06,
+      "loss": 0.1366,
+      "step": 1234
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.84870701146361e-06,
+      "loss": 0.1624,
+      "step": 1235
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.848573713676353e-06,
+      "loss": 0.1705,
+      "step": 1236
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.848440415889096e-06,
+      "loss": 0.1357,
+      "step": 1237
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.84830711810184e-06,
+      "loss": 0.1097,
+      "step": 1238
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.848173820314583e-06,
+      "loss": 0.149,
+      "step": 1239
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.848040522527326e-06,
+      "loss": 0.1186,
+      "step": 1240
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.84790722474007e-06,
+      "loss": 0.1092,
+      "step": 1241
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.847773926952813e-06,
+      "loss": 0.1508,
+      "step": 1242
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.847640629165556e-06,
+      "loss": 0.168,
+      "step": 1243
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.847507331378299e-06,
+      "loss": 0.2179,
+      "step": 1244
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.847374033591044e-06,
+      "loss": 0.1226,
+      "step": 1245
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.847240735803786e-06,
+      "loss": 0.1152,
+      "step": 1246
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.84710743801653e-06,
+      "loss": 0.1134,
+      "step": 1247
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.846974140229274e-06,
+      "loss": 0.1271,
+      "step": 1248
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.846840842442017e-06,
+      "loss": 0.1748,
+      "step": 1249
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.84670754465476e-06,
+      "loss": 0.171,
+      "step": 1250
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.846574246867502e-06,
+      "loss": 0.1708,
+      "step": 1251
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.846440949080247e-06,
+      "loss": 0.1494,
+      "step": 1252
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.84630765129299e-06,
+      "loss": 0.1298,
+      "step": 1253
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.846174353505732e-06,
+      "loss": 0.1634,
+      "step": 1254
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.846041055718477e-06,
+      "loss": 0.1409,
+      "step": 1255
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.84590775793122e-06,
+      "loss": 0.1239,
+      "step": 1256
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.845774460143962e-06,
+      "loss": 0.1435,
+      "step": 1257
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.845641162356705e-06,
+      "loss": 0.1456,
+      "step": 1258
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.84550786456945e-06,
+      "loss": 0.1657,
+      "step": 1259
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.845374566782193e-06,
+      "loss": 0.149,
+      "step": 1260
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.845241268994935e-06,
+      "loss": 0.1228,
+      "step": 1261
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.84510797120768e-06,
+      "loss": 0.1432,
+      "step": 1262
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.844974673420423e-06,
+      "loss": 0.1437,
+      "step": 1263
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.844841375633166e-06,
+      "loss": 0.1698,
+      "step": 1264
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.844708077845908e-06,
+      "loss": 0.1397,
+      "step": 1265
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.844574780058653e-06,
+      "loss": 0.1638,
+      "step": 1266
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.844441482271396e-06,
+      "loss": 0.1684,
+      "step": 1267
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.844308184484139e-06,
+      "loss": 0.1382,
+      "step": 1268
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.844174886696881e-06,
+      "loss": 0.1315,
+      "step": 1269
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.844041588909626e-06,
+      "loss": 0.1407,
+      "step": 1270
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.843908291122369e-06,
+      "loss": 0.2015,
+      "step": 1271
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.843774993335112e-06,
+      "loss": 0.2166,
+      "step": 1272
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.843641695547854e-06,
+      "loss": 0.1576,
+      "step": 1273
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.843508397760599e-06,
+      "loss": 0.1552,
+      "step": 1274
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.843375099973342e-06,
+      "loss": 0.0959,
+      "step": 1275
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.843241802186085e-06,
+      "loss": 0.1151,
+      "step": 1276
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.843108504398827e-06,
+      "loss": 0.155,
+      "step": 1277
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.842975206611572e-06,
+      "loss": 0.1543,
+      "step": 1278
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.842841908824315e-06,
+      "loss": 0.1537,
+      "step": 1279
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.842708611037057e-06,
+      "loss": 0.1683,
+      "step": 1280
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.8425753132498e-06,
+      "loss": 0.1513,
+      "step": 1281
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.842442015462545e-06,
+      "loss": 0.1458,
+      "step": 1282
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.842308717675288e-06,
+      "loss": 0.1487,
+      "step": 1283
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.84217541988803e-06,
+      "loss": 0.1414,
+      "step": 1284
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.842042122100775e-06,
+      "loss": 0.1622,
+      "step": 1285
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.841908824313518e-06,
+      "loss": 0.1318,
+      "step": 1286
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.84177552652626e-06,
+      "loss": 0.0979,
+      "step": 1287
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.841642228739003e-06,
+      "loss": 0.1429,
+      "step": 1288
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.841508930951748e-06,
+      "loss": 0.1356,
+      "step": 1289
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.84137563316449e-06,
+      "loss": 0.1025,
+      "step": 1290
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.841242335377234e-06,
+      "loss": 0.1603,
+      "step": 1291
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.841109037589976e-06,
+      "loss": 0.1569,
+      "step": 1292
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.840975739802721e-06,
+      "loss": 0.1443,
+      "step": 1293
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.840842442015464e-06,
+      "loss": 0.1304,
+      "step": 1294
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.840709144228207e-06,
+      "loss": 0.1415,
+      "step": 1295
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.84057584644095e-06,
+      "loss": 0.1522,
+      "step": 1296
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.840442548653694e-06,
+      "loss": 0.1168,
+      "step": 1297
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.840309250866437e-06,
+      "loss": 0.1443,
+      "step": 1298
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.84017595307918e-06,
+      "loss": 0.178,
+      "step": 1299
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.840042655291922e-06,
+      "loss": 0.2247,
+      "step": 1300
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.839909357504667e-06,
+      "loss": 0.1626,
+      "step": 1301
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.83977605971741e-06,
+      "loss": 0.1468,
+      "step": 1302
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.839642761930152e-06,
+      "loss": 0.0967,
+      "step": 1303
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.839509464142895e-06,
+      "loss": 0.1152,
+      "step": 1304
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.83937616635564e-06,
+      "loss": 0.1769,
+      "step": 1305
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.839242868568383e-06,
+      "loss": 0.1338,
+      "step": 1306
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.839109570781125e-06,
+      "loss": 0.1554,
+      "step": 1307
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.838976272993868e-06,
+      "loss": 0.1667,
+      "step": 1308
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.838842975206613e-06,
+      "loss": 0.122,
+      "step": 1309
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.838709677419356e-06,
+      "loss": 0.16,
+      "step": 1310
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.838576379632098e-06,
+      "loss": 0.1302,
+      "step": 1311
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.838443081844841e-06,
+      "loss": 0.1379,
+      "step": 1312
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.838309784057586e-06,
+      "loss": 0.1398,
+      "step": 1313
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.838176486270329e-06,
+      "loss": 0.1267,
+      "step": 1314
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.838043188483071e-06,
+      "loss": 0.1651,
+      "step": 1315
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.837909890695814e-06,
+      "loss": 0.1082,
+      "step": 1316
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.837776592908559e-06,
+      "loss": 0.1274,
+      "step": 1317
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.837643295121302e-06,
+      "loss": 0.1604,
+      "step": 1318
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.837509997334044e-06,
+      "loss": 0.1613,
+      "step": 1319
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.837376699546787e-06,
+      "loss": 0.098,
+      "step": 1320
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.837243401759532e-06,
+      "loss": 0.1148,
+      "step": 1321
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.837110103972275e-06,
+      "loss": 0.1866,
+      "step": 1322
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.836976806185017e-06,
+      "loss": 0.191,
+      "step": 1323
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.83684350839776e-06,
+      "loss": 0.1493,
+      "step": 1324
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.836710210610505e-06,
+      "loss": 0.1271,
+      "step": 1325
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.836576912823247e-06,
+      "loss": 0.135,
+      "step": 1326
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.83644361503599e-06,
+      "loss": 0.1167,
+      "step": 1327
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.836310317248733e-06,
+      "loss": 0.1408,
+      "step": 1328
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.836177019461478e-06,
+      "loss": 0.1655,
+      "step": 1329
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.83604372167422e-06,
+      "loss": 0.1438,
+      "step": 1330
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.835910423886963e-06,
+      "loss": 0.218,
+      "step": 1331
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.835777126099708e-06,
+      "loss": 0.1425,
+      "step": 1332
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.83564382831245e-06,
+      "loss": 0.1564,
+      "step": 1333
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.835510530525193e-06,
+      "loss": 0.1394,
+      "step": 1334
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.835377232737936e-06,
+      "loss": 0.1369,
+      "step": 1335
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.83524393495068e-06,
+      "loss": 0.1384,
+      "step": 1336
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.835110637163424e-06,
+      "loss": 0.1463,
+      "step": 1337
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.834977339376166e-06,
+      "loss": 0.1379,
+      "step": 1338
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.834844041588911e-06,
+      "loss": 0.1415,
+      "step": 1339
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.834710743801654e-06,
+      "loss": 0.1824,
+      "step": 1340
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.834577446014397e-06,
+      "loss": 0.1994,
+      "step": 1341
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.83444414822714e-06,
+      "loss": 0.149,
+      "step": 1342
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.834310850439884e-06,
+      "loss": 0.1501,
+      "step": 1343
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.834177552652627e-06,
+      "loss": 0.1468,
+      "step": 1344
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.83404425486537e-06,
+      "loss": 0.1843,
+      "step": 1345
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.833910957078114e-06,
+      "loss": 0.1689,
+      "step": 1346
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.833777659290857e-06,
+      "loss": 0.097,
+      "step": 1347
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.8336443615036e-06,
+      "loss": 0.1179,
+      "step": 1348
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.833511063716342e-06,
+      "loss": 0.1773,
+      "step": 1349
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.833377765929087e-06,
+      "loss": 0.111,
+      "step": 1350
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.83324446814183e-06,
+      "loss": 0.1147,
+      "step": 1351
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.833111170354573e-06,
+      "loss": 0.1492,
+      "step": 1352
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.832977872567317e-06,
+      "loss": 0.1335,
+      "step": 1353
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.83284457478006e-06,
+      "loss": 0.1356,
+      "step": 1354
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.832711276992803e-06,
+      "loss": 0.1394,
+      "step": 1355
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.832577979205546e-06,
+      "loss": 0.1976,
+      "step": 1356
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.83244468141829e-06,
+      "loss": 0.1188,
+      "step": 1357
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.832311383631033e-06,
+      "loss": 0.1266,
+      "step": 1358
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.832178085843776e-06,
+      "loss": 0.1649,
+      "step": 1359
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.83204478805652e-06,
+      "loss": 0.1351,
+      "step": 1360
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.831911490269263e-06,
+      "loss": 0.1883,
+      "step": 1361
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.831778192482006e-06,
+      "loss": 0.1735,
+      "step": 1362
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.831644894694749e-06,
+      "loss": 0.1577,
+      "step": 1363
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.831511596907493e-06,
+      "loss": 0.1288,
+      "step": 1364
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.831378299120236e-06,
+      "loss": 0.1658,
+      "step": 1365
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.831245001332979e-06,
+      "loss": 0.1476,
+      "step": 1366
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.831111703545722e-06,
+      "loss": 0.1014,
+      "step": 1367
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.830978405758466e-06,
+      "loss": 0.1267,
+      "step": 1368
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.830845107971209e-06,
+      "loss": 0.1506,
+      "step": 1369
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.830711810183952e-06,
+      "loss": 0.1443,
+      "step": 1370
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.830578512396695e-06,
+      "loss": 0.1313,
+      "step": 1371
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.83044521460944e-06,
+      "loss": 0.1227,
+      "step": 1372
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.830311916822182e-06,
+      "loss": 0.1404,
+      "step": 1373
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.830178619034925e-06,
+      "loss": 0.1297,
+      "step": 1374
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.830045321247668e-06,
+      "loss": 0.1788,
+      "step": 1375
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.829912023460412e-06,
+      "loss": 0.164,
+      "step": 1376
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.829778725673155e-06,
+      "loss": 0.117,
+      "step": 1377
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.829645427885898e-06,
+      "loss": 0.1179,
+      "step": 1378
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.82951213009864e-06,
+      "loss": 0.139,
+      "step": 1379
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.829378832311385e-06,
+      "loss": 0.2051,
+      "step": 1380
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.829245534524128e-06,
+      "loss": 0.1343,
+      "step": 1381
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.82911223673687e-06,
+      "loss": 0.1447,
+      "step": 1382
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.828978938949614e-06,
+      "loss": 0.1441,
+      "step": 1383
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.828845641162358e-06,
+      "loss": 0.1306,
+      "step": 1384
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.828712343375101e-06,
+      "loss": 0.1501,
+      "step": 1385
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.828579045587844e-06,
+      "loss": 0.1437,
+      "step": 1386
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.828445747800587e-06,
+      "loss": 0.1449,
+      "step": 1387
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.828312450013331e-06,
+      "loss": 0.1147,
+      "step": 1388
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.828179152226074e-06,
+      "loss": 0.1477,
+      "step": 1389
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.828045854438817e-06,
+      "loss": 0.1484,
+      "step": 1390
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.82791255665156e-06,
+      "loss": 0.1157,
+      "step": 1391
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.827779258864304e-06,
+      "loss": 0.1364,
+      "step": 1392
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.827645961077047e-06,
+      "loss": 0.1453,
+      "step": 1393
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.82751266328979e-06,
+      "loss": 0.1527,
+      "step": 1394
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.827379365502533e-06,
+      "loss": 0.1373,
+      "step": 1395
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.827246067715277e-06,
+      "loss": 0.131,
+      "step": 1396
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.82711276992802e-06,
+      "loss": 0.1874,
+      "step": 1397
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.826979472140763e-06,
+      "loss": 0.1285,
+      "step": 1398
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.826846174353507e-06,
+      "loss": 0.1975,
+      "step": 1399
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.82671287656625e-06,
+      "loss": 0.1625,
+      "step": 1400
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.826579578778993e-06,
+      "loss": 0.2216,
+      "step": 1401
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.826446280991736e-06,
+      "loss": 0.1442,
+      "step": 1402
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.82631298320448e-06,
+      "loss": 0.121,
+      "step": 1403
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.826179685417223e-06,
+      "loss": 0.1845,
+      "step": 1404
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.826046387629966e-06,
+      "loss": 0.1393,
+      "step": 1405
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.825913089842709e-06,
+      "loss": 0.1302,
+      "step": 1406
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.825779792055453e-06,
+      "loss": 0.1867,
+      "step": 1407
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.825646494268196e-06,
+      "loss": 0.1665,
+      "step": 1408
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.825513196480939e-06,
+      "loss": 0.1191,
+      "step": 1409
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.825379898693682e-06,
+      "loss": 0.1181,
+      "step": 1410
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.825246600906426e-06,
+      "loss": 0.1961,
+      "step": 1411
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.825113303119169e-06,
+      "loss": 0.1187,
+      "step": 1412
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.824980005331912e-06,
+      "loss": 0.1631,
+      "step": 1413
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.824846707544655e-06,
+      "loss": 0.1239,
+      "step": 1414
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.824713409757399e-06,
+      "loss": 0.1156,
+      "step": 1415
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.824580111970142e-06,
+      "loss": 0.1391,
+      "step": 1416
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.824446814182885e-06,
+      "loss": 0.1536,
+      "step": 1417
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.824313516395628e-06,
+      "loss": 0.1198,
+      "step": 1418
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.824180218608372e-06,
+      "loss": 0.1751,
+      "step": 1419
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.824046920821115e-06,
+      "loss": 0.1563,
+      "step": 1420
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.823913623033858e-06,
+      "loss": 0.1035,
+      "step": 1421
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.8237803252466e-06,
+      "loss": 0.1358,
+      "step": 1422
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.823647027459345e-06,
+      "loss": 0.1284,
+      "step": 1423
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.823513729672088e-06,
+      "loss": 0.1242,
+      "step": 1424
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.82338043188483e-06,
+      "loss": 0.165,
+      "step": 1425
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.823247134097575e-06,
+      "loss": 0.107,
+      "step": 1426
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.823113836310318e-06,
+      "loss": 0.163,
+      "step": 1427
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.82298053852306e-06,
+      "loss": 0.1406,
+      "step": 1428
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.822847240735804e-06,
+      "loss": 0.1245,
+      "step": 1429
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.822713942948548e-06,
+      "loss": 0.126,
+      "step": 1430
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.822580645161291e-06,
+      "loss": 0.1406,
+      "step": 1431
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.822447347374034e-06,
+      "loss": 0.1518,
+      "step": 1432
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.822314049586778e-06,
+      "loss": 0.1498,
+      "step": 1433
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.822180751799521e-06,
+      "loss": 0.1804,
+      "step": 1434
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.822047454012264e-06,
+      "loss": 0.1049,
+      "step": 1435
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.821914156225007e-06,
+      "loss": 0.1829,
+      "step": 1436
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.821780858437751e-06,
+      "loss": 0.1576,
+      "step": 1437
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.821647560650494e-06,
+      "loss": 0.1616,
+      "step": 1438
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.821514262863237e-06,
+      "loss": 0.1432,
+      "step": 1439
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.821380965075981e-06,
+      "loss": 0.1437,
+      "step": 1440
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.821247667288724e-06,
+      "loss": 0.1554,
+      "step": 1441
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.821114369501467e-06,
+      "loss": 0.1769,
+      "step": 1442
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.82098107171421e-06,
+      "loss": 0.1379,
+      "step": 1443
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.820847773926954e-06,
+      "loss": 0.1534,
+      "step": 1444
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.820714476139697e-06,
+      "loss": 0.1527,
+      "step": 1445
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.82058117835244e-06,
+      "loss": 0.1452,
+      "step": 1446
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.820447880565185e-06,
+      "loss": 0.1046,
+      "step": 1447
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.820314582777927e-06,
+      "loss": 0.1395,
+      "step": 1448
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.82018128499067e-06,
+      "loss": 0.1899,
+      "step": 1449
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.820047987203413e-06,
+      "loss": 0.1275,
+      "step": 1450
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.819914689416157e-06,
+      "loss": 0.1513,
+      "step": 1451
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.8197813916289e-06,
+      "loss": 0.1013,
+      "step": 1452
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.819648093841643e-06,
+      "loss": 0.1444,
+      "step": 1453
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.819514796054388e-06,
+      "loss": 0.1307,
+      "step": 1454
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.81938149826713e-06,
+      "loss": 0.0958,
+      "step": 1455
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.819248200479873e-06,
+      "loss": 0.128,
+      "step": 1456
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.819114902692616e-06,
+      "loss": 0.191,
+      "step": 1457
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.81898160490536e-06,
+      "loss": 0.1289,
+      "step": 1458
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.818848307118103e-06,
+      "loss": 0.1145,
+      "step": 1459
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.818715009330846e-06,
+      "loss": 0.1153,
+      "step": 1460
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.818581711543589e-06,
+      "loss": 0.1446,
+      "step": 1461
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.818448413756334e-06,
+      "loss": 0.1664,
+      "step": 1462
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.818315115969076e-06,
+      "loss": 0.115,
+      "step": 1463
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.81818181818182e-06,
+      "loss": 0.1544,
+      "step": 1464
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.818048520394562e-06,
+      "loss": 0.1226,
+      "step": 1465
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.817915222607307e-06,
+      "loss": 0.1252,
+      "step": 1466
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.81778192482005e-06,
+      "loss": 0.1397,
+      "step": 1467
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.817648627032792e-06,
+      "loss": 0.1209,
+      "step": 1468
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.817515329245535e-06,
+      "loss": 0.1615,
+      "step": 1469
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.81738203145828e-06,
+      "loss": 0.1518,
+      "step": 1470
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.817248733671022e-06,
+      "loss": 0.1117,
+      "step": 1471
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.817115435883765e-06,
+      "loss": 0.1666,
+      "step": 1472
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.816982138096508e-06,
+      "loss": 0.146,
+      "step": 1473
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.816848840309252e-06,
+      "loss": 0.1318,
+      "step": 1474
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.816715542521995e-06,
+      "loss": 0.1566,
+      "step": 1475
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.816582244734738e-06,
+      "loss": 0.1113,
+      "step": 1476
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.816448946947481e-06,
+      "loss": 0.1291,
+      "step": 1477
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.816315649160225e-06,
+      "loss": 0.1657,
+      "step": 1478
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.816182351372968e-06,
+      "loss": 0.2098,
+      "step": 1479
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.816049053585711e-06,
+      "loss": 0.1511,
+      "step": 1480
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.815915755798454e-06,
+      "loss": 0.1592,
+      "step": 1481
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.815782458011198e-06,
+      "loss": 0.1497,
+      "step": 1482
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.815649160223941e-06,
+      "loss": 0.1518,
+      "step": 1483
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.815515862436684e-06,
+      "loss": 0.1462,
+      "step": 1484
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.815382564649427e-06,
+      "loss": 0.1473,
+      "step": 1485
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.815249266862171e-06,
+      "loss": 0.1468,
+      "step": 1486
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.815115969074914e-06,
+      "loss": 0.118,
+      "step": 1487
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.814982671287657e-06,
+      "loss": 0.111,
+      "step": 1488
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.8148493735004e-06,
+      "loss": 0.1452,
+      "step": 1489
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.814716075713144e-06,
+      "loss": 0.1109,
+      "step": 1490
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.814582777925887e-06,
+      "loss": 0.1761,
+      "step": 1491
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.81444948013863e-06,
+      "loss": 0.1219,
+      "step": 1492
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.814316182351373e-06,
+      "loss": 0.1749,
+      "step": 1493
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.814182884564117e-06,
+      "loss": 0.1278,
+      "step": 1494
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.81404958677686e-06,
+      "loss": 0.1392,
+      "step": 1495
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.813916288989603e-06,
+      "loss": 0.216,
+      "step": 1496
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.813782991202346e-06,
+      "loss": 0.1389,
+      "step": 1497
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.81364969341509e-06,
+      "loss": 0.1542,
+      "step": 1498
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.813516395627833e-06,
+      "loss": 0.1655,
+      "step": 1499
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.813383097840576e-06,
+      "loss": 0.158,
+      "step": 1500
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.813249800053319e-06,
+      "loss": 0.1367,
+      "step": 1501
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.813116502266063e-06,
+      "loss": 0.1287,
+      "step": 1502
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.812983204478806e-06,
+      "loss": 0.1525,
+      "step": 1503
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.812849906691549e-06,
+      "loss": 0.1405,
+      "step": 1504
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.812716608904292e-06,
+      "loss": 0.1071,
+      "step": 1505
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.812583311117036e-06,
+      "loss": 0.1519,
+      "step": 1506
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.812450013329779e-06,
+      "loss": 0.1303,
+      "step": 1507
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.812316715542522e-06,
+      "loss": 0.1194,
+      "step": 1508
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.812183417755266e-06,
+      "loss": 0.1356,
+      "step": 1509
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.81205011996801e-06,
+      "loss": 0.1597,
+      "step": 1510
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.811916822180752e-06,
+      "loss": 0.129,
+      "step": 1511
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.811783524393495e-06,
+      "loss": 0.1333,
+      "step": 1512
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.81165022660624e-06,
+      "loss": 0.1241,
+      "step": 1513
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.811516928818982e-06,
+      "loss": 0.1581,
+      "step": 1514
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.811383631031725e-06,
+      "loss": 0.1279,
+      "step": 1515
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.811250333244468e-06,
+      "loss": 0.1723,
+      "step": 1516
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.811117035457212e-06,
+      "loss": 0.2155,
+      "step": 1517
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.810983737669955e-06,
+      "loss": 0.1333,
+      "step": 1518
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.810850439882698e-06,
+      "loss": 0.1721,
+      "step": 1519
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.810717142095442e-06,
+      "loss": 0.1272,
+      "step": 1520
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.810583844308185e-06,
+      "loss": 0.1474,
+      "step": 1521
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.810450546520928e-06,
+      "loss": 0.1538,
+      "step": 1522
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.810317248733671e-06,
+      "loss": 0.1513,
+      "step": 1523
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.810183950946415e-06,
+      "loss": 0.1312,
+      "step": 1524
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.810050653159158e-06,
+      "loss": 0.1216,
+      "step": 1525
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.809917355371901e-06,
+      "loss": 0.143,
+      "step": 1526
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.809784057584646e-06,
+      "loss": 0.1626,
+      "step": 1527
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.809650759797388e-06,
+      "loss": 0.1183,
+      "step": 1528
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.809517462010131e-06,
+      "loss": 0.1417,
+      "step": 1529
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.809384164222874e-06,
+      "loss": 0.1144,
+      "step": 1530
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.809250866435619e-06,
+      "loss": 0.1325,
+      "step": 1531
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.809117568648361e-06,
+      "loss": 0.1181,
+      "step": 1532
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.808984270861104e-06,
+      "loss": 0.0977,
+      "step": 1533
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.808850973073849e-06,
+      "loss": 0.1595,
+      "step": 1534
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.808717675286592e-06,
+      "loss": 0.1598,
+      "step": 1535
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.808584377499334e-06,
+      "loss": 0.1638,
+      "step": 1536
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.808451079712077e-06,
+      "loss": 0.0842,
+      "step": 1537
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.808317781924822e-06,
+      "loss": 0.1765,
+      "step": 1538
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.808184484137565e-06,
+      "loss": 0.1466,
+      "step": 1539
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.808051186350307e-06,
+      "loss": 0.1186,
+      "step": 1540
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.807917888563052e-06,
+      "loss": 0.1364,
+      "step": 1541
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.807784590775795e-06,
+      "loss": 0.1395,
+      "step": 1542
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.807651292988537e-06,
+      "loss": 0.172,
+      "step": 1543
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.80751799520128e-06,
+      "loss": 0.1899,
+      "step": 1544
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.807384697414025e-06,
+      "loss": 0.1408,
+      "step": 1545
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.807251399626768e-06,
+      "loss": 0.148,
+      "step": 1546
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.80711810183951e-06,
+      "loss": 0.2109,
+      "step": 1547
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.806984804052253e-06,
+      "loss": 0.1319,
+      "step": 1548
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.806851506264998e-06,
+      "loss": 0.1039,
+      "step": 1549
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.80671820847774e-06,
+      "loss": 0.1406,
+      "step": 1550
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.806584910690483e-06,
+      "loss": 0.1135,
+      "step": 1551
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.806451612903226e-06,
+      "loss": 0.1363,
+      "step": 1552
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.80631831511597e-06,
+      "loss": 0.096,
+      "step": 1553
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.806185017328714e-06,
+      "loss": 0.1893,
+      "step": 1554
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.806051719541456e-06,
+      "loss": 0.1662,
+      "step": 1555
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.8059184217542e-06,
+      "loss": 0.1331,
+      "step": 1556
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.805785123966944e-06,
+      "loss": 0.1287,
+      "step": 1557
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.805651826179687e-06,
+      "loss": 0.1048,
+      "step": 1558
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.80551852839243e-06,
+      "loss": 0.1321,
+      "step": 1559
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.805385230605172e-06,
+      "loss": 0.1536,
+      "step": 1560
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.805251932817917e-06,
+      "loss": 0.1155,
+      "step": 1561
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.80511863503066e-06,
+      "loss": 0.1313,
+      "step": 1562
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.804985337243402e-06,
+      "loss": 0.12,
+      "step": 1563
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.804852039456145e-06,
+      "loss": 0.1154,
+      "step": 1564
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.80471874166889e-06,
+      "loss": 0.1153,
+      "step": 1565
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.804585443881632e-06,
+      "loss": 0.1384,
+      "step": 1566
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.804452146094375e-06,
+      "loss": 0.1785,
+      "step": 1567
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.80431884830712e-06,
+      "loss": 0.1313,
+      "step": 1568
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.804185550519863e-06,
+      "loss": 0.1414,
+      "step": 1569
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.804052252732605e-06,
+      "loss": 0.2052,
+      "step": 1570
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.803918954945348e-06,
+      "loss": 0.1409,
+      "step": 1571
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.803785657158093e-06,
+      "loss": 0.1208,
+      "step": 1572
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.803652359370836e-06,
+      "loss": 0.1406,
+      "step": 1573
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.803519061583578e-06,
+      "loss": 0.1629,
+      "step": 1574
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.803385763796321e-06,
+      "loss": 0.1563,
+      "step": 1575
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.803252466009066e-06,
+      "loss": 0.1592,
+      "step": 1576
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.803119168221809e-06,
+      "loss": 0.0897,
+      "step": 1577
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.802985870434551e-06,
+      "loss": 0.1139,
+      "step": 1578
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.802852572647294e-06,
+      "loss": 0.1311,
+      "step": 1579
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.802719274860039e-06,
+      "loss": 0.1239,
+      "step": 1580
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.802585977072782e-06,
+      "loss": 0.1483,
+      "step": 1581
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.802452679285524e-06,
+      "loss": 0.1473,
+      "step": 1582
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.802319381498267e-06,
+      "loss": 0.1188,
+      "step": 1583
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.802186083711012e-06,
+      "loss": 0.1188,
+      "step": 1584
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.802052785923755e-06,
+      "loss": 0.152,
+      "step": 1585
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.801919488136497e-06,
+      "loss": 0.1287,
+      "step": 1586
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.80178619034924e-06,
+      "loss": 0.1736,
+      "step": 1587
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.801652892561985e-06,
+      "loss": 0.1341,
+      "step": 1588
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.801519594774727e-06,
+      "loss": 0.1133,
+      "step": 1589
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.80138629698747e-06,
+      "loss": 0.1328,
+      "step": 1590
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.801252999200213e-06,
+      "loss": 0.1319,
+      "step": 1591
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.801119701412958e-06,
+      "loss": 0.1637,
+      "step": 1592
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.8009864036257e-06,
+      "loss": 0.1449,
+      "step": 1593
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.800853105838443e-06,
+      "loss": 0.1446,
+      "step": 1594
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.800719808051186e-06,
+      "loss": 0.0939,
+      "step": 1595
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.80058651026393e-06,
+      "loss": 0.1674,
+      "step": 1596
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.800453212476673e-06,
+      "loss": 0.1422,
+      "step": 1597
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.800319914689416e-06,
+      "loss": 0.1544,
+      "step": 1598
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.800186616902159e-06,
+      "loss": 0.1548,
+      "step": 1599
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.800053319114904e-06,
+      "loss": 0.1698,
+      "step": 1600
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.799920021327646e-06,
+      "loss": 0.1414,
+      "step": 1601
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.79978672354039e-06,
+      "loss": 0.1149,
+      "step": 1602
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.799653425753132e-06,
+      "loss": 0.1485,
+      "step": 1603
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.799520127965877e-06,
+      "loss": 0.1259,
+      "step": 1604
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.79938683017862e-06,
+      "loss": 0.1121,
+      "step": 1605
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.799253532391362e-06,
+      "loss": 0.1419,
+      "step": 1606
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.799120234604107e-06,
+      "loss": 0.147,
+      "step": 1607
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.79898693681685e-06,
+      "loss": 0.0957,
+      "step": 1608
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.798853639029592e-06,
+      "loss": 0.1392,
+      "step": 1609
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.798720341242335e-06,
+      "loss": 0.1481,
+      "step": 1610
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.79858704345508e-06,
+      "loss": 0.1757,
+      "step": 1611
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.798453745667822e-06,
+      "loss": 0.1667,
+      "step": 1612
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.798320447880565e-06,
+      "loss": 0.143,
+      "step": 1613
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.79818715009331e-06,
+      "loss": 0.1292,
+      "step": 1614
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.798053852306053e-06,
+      "loss": 0.1445,
+      "step": 1615
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.797920554518795e-06,
+      "loss": 0.1322,
+      "step": 1616
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.797787256731538e-06,
+      "loss": 0.0963,
+      "step": 1617
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.797653958944283e-06,
+      "loss": 0.1444,
+      "step": 1618
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.797520661157026e-06,
+      "loss": 0.1227,
+      "step": 1619
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.797387363369768e-06,
+      "loss": 0.12,
+      "step": 1620
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.797254065582513e-06,
+      "loss": 0.1356,
+      "step": 1621
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.797120767795256e-06,
+      "loss": 0.1424,
+      "step": 1622
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.796987470007999e-06,
+      "loss": 0.159,
+      "step": 1623
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.796854172220741e-06,
+      "loss": 0.1148,
+      "step": 1624
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.796720874433486e-06,
+      "loss": 0.1069,
+      "step": 1625
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.796587576646229e-06,
+      "loss": 0.1265,
+      "step": 1626
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.796454278858972e-06,
+      "loss": 0.1181,
+      "step": 1627
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.796320981071716e-06,
+      "loss": 0.1216,
+      "step": 1628
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.796187683284459e-06,
+      "loss": 0.1486,
+      "step": 1629
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.796054385497202e-06,
+      "loss": 0.1136,
+      "step": 1630
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.795921087709945e-06,
+      "loss": 0.1384,
+      "step": 1631
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.795787789922689e-06,
+      "loss": 0.1474,
+      "step": 1632
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.795654492135432e-06,
+      "loss": 0.1579,
+      "step": 1633
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.795521194348175e-06,
+      "loss": 0.1364,
+      "step": 1634
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.79538789656092e-06,
+      "loss": 0.1174,
+      "step": 1635
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.795254598773662e-06,
+      "loss": 0.1488,
+      "step": 1636
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.795121300986405e-06,
+      "loss": 0.1453,
+      "step": 1637
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.794988003199148e-06,
+      "loss": 0.159,
+      "step": 1638
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.794854705411892e-06,
+      "loss": 0.1027,
+      "step": 1639
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.794721407624635e-06,
+      "loss": 0.1115,
+      "step": 1640
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.794588109837378e-06,
+      "loss": 0.1476,
+      "step": 1641
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.79445481205012e-06,
+      "loss": 0.1531,
+      "step": 1642
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.794321514262865e-06,
+      "loss": 0.1294,
+      "step": 1643
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.794188216475608e-06,
+      "loss": 0.1569,
+      "step": 1644
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.79405491868835e-06,
+      "loss": 0.1246,
+      "step": 1645
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.793921620901094e-06,
+      "loss": 0.1588,
+      "step": 1646
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.793788323113838e-06,
+      "loss": 0.146,
+      "step": 1647
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.793655025326581e-06,
+      "loss": 0.1474,
+      "step": 1648
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.793521727539324e-06,
+      "loss": 0.1509,
+      "step": 1649
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.793388429752067e-06,
+      "loss": 0.0878,
+      "step": 1650
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.793255131964811e-06,
+      "loss": 0.1768,
+      "step": 1651
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.793121834177554e-06,
+      "loss": 0.0965,
+      "step": 1652
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.792988536390297e-06,
+      "loss": 0.1104,
+      "step": 1653
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.79285523860304e-06,
+      "loss": 0.1612,
+      "step": 1654
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.792721940815784e-06,
+      "loss": 0.1361,
+      "step": 1655
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.792588643028527e-06,
+      "loss": 0.1401,
+      "step": 1656
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.79245534524127e-06,
+      "loss": 0.1355,
+      "step": 1657
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.792322047454012e-06,
+      "loss": 0.1558,
+      "step": 1658
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.792188749666757e-06,
+      "loss": 0.1973,
+      "step": 1659
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.7920554518795e-06,
+      "loss": 0.1393,
+      "step": 1660
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.791922154092243e-06,
+      "loss": 0.1168,
+      "step": 1661
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.791788856304985e-06,
+      "loss": 0.1036,
+      "step": 1662
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.79165555851773e-06,
+      "loss": 0.164,
+      "step": 1663
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.791522260730473e-06,
+      "loss": 0.1338,
+      "step": 1664
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.791388962943216e-06,
+      "loss": 0.1511,
+      "step": 1665
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.791255665155958e-06,
+      "loss": 0.1175,
+      "step": 1666
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.791122367368703e-06,
+      "loss": 0.1763,
+      "step": 1667
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.790989069581446e-06,
+      "loss": 0.1217,
+      "step": 1668
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.790855771794189e-06,
+      "loss": 0.1055,
+      "step": 1669
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.790722474006931e-06,
+      "loss": 0.1193,
+      "step": 1670
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.790589176219676e-06,
+      "loss": 0.1046,
+      "step": 1671
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.790455878432419e-06,
+      "loss": 0.1131,
+      "step": 1672
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.790322580645162e-06,
+      "loss": 0.1414,
+      "step": 1673
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.790189282857904e-06,
+      "loss": 0.1326,
+      "step": 1674
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.790055985070649e-06,
+      "loss": 0.1659,
+      "step": 1675
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.789922687283392e-06,
+      "loss": 0.0931,
+      "step": 1676
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.789789389496135e-06,
+      "loss": 0.1283,
+      "step": 1677
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.789656091708879e-06,
+      "loss": 0.1311,
+      "step": 1678
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.789522793921622e-06,
+      "loss": 0.1428,
+      "step": 1679
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.789389496134365e-06,
+      "loss": 0.1506,
+      "step": 1680
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.789256198347107e-06,
+      "loss": 0.16,
+      "step": 1681
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.789122900559852e-06,
+      "loss": 0.1282,
+      "step": 1682
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.788989602772595e-06,
+      "loss": 0.0957,
+      "step": 1683
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.788856304985338e-06,
+      "loss": 0.1457,
+      "step": 1684
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.78872300719808e-06,
+      "loss": 0.1581,
+      "step": 1685
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.788589709410825e-06,
+      "loss": 0.1442,
+      "step": 1686
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.788456411623568e-06,
+      "loss": 0.1172,
+      "step": 1687
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.78832311383631e-06,
+      "loss": 0.1446,
+      "step": 1688
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.788189816049053e-06,
+      "loss": 0.0777,
+      "step": 1689
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.788056518261798e-06,
+      "loss": 0.17,
+      "step": 1690
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.78792322047454e-06,
+      "loss": 0.1509,
+      "step": 1691
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.787789922687284e-06,
+      "loss": 0.108,
+      "step": 1692
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.787656624900026e-06,
+      "loss": 0.1631,
+      "step": 1693
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.787523327112771e-06,
+      "loss": 0.1112,
+      "step": 1694
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.787390029325514e-06,
+      "loss": 0.136,
+      "step": 1695
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.787256731538257e-06,
+      "loss": 0.1447,
+      "step": 1696
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.787123433751e-06,
+      "loss": 0.1584,
+      "step": 1697
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.786990135963744e-06,
+      "loss": 0.1452,
+      "step": 1698
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.786856838176487e-06,
+      "loss": 0.1256,
+      "step": 1699
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.78672354038923e-06,
+      "loss": 0.1198,
+      "step": 1700
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.786590242601974e-06,
+      "loss": 0.1098,
+      "step": 1701
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.786456944814717e-06,
+      "loss": 0.1539,
+      "step": 1702
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.78632364702746e-06,
+      "loss": 0.138,
+      "step": 1703
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.786190349240202e-06,
+      "loss": 0.138,
+      "step": 1704
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.786057051452947e-06,
+      "loss": 0.1094,
+      "step": 1705
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.78592375366569e-06,
+      "loss": 0.143,
+      "step": 1706
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.785790455878433e-06,
+      "loss": 0.1189,
+      "step": 1707
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.785657158091177e-06,
+      "loss": 0.1411,
+      "step": 1708
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.78552386030392e-06,
+      "loss": 0.129,
+      "step": 1709
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.785390562516663e-06,
+      "loss": 0.1183,
+      "step": 1710
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.785257264729406e-06,
+      "loss": 0.1298,
+      "step": 1711
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.78512396694215e-06,
+      "loss": 0.1565,
+      "step": 1712
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.784990669154893e-06,
+      "loss": 0.1159,
+      "step": 1713
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.784857371367636e-06,
+      "loss": 0.1205,
+      "step": 1714
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.78472407358038e-06,
+      "loss": 0.1759,
+      "step": 1715
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.784590775793123e-06,
+      "loss": 0.158,
+      "step": 1716
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.784457478005866e-06,
+      "loss": 0.1755,
+      "step": 1717
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.784324180218609e-06,
+      "loss": 0.1219,
+      "step": 1718
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.784190882431353e-06,
+      "loss": 0.1663,
+      "step": 1719
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.784057584644096e-06,
+      "loss": 0.1365,
+      "step": 1720
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.783924286856839e-06,
+      "loss": 0.1323,
+      "step": 1721
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.783790989069583e-06,
+      "loss": 0.1096,
+      "step": 1722
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.783657691282326e-06,
+      "loss": 0.1769,
+      "step": 1723
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.783524393495069e-06,
+      "loss": 0.1117,
+      "step": 1724
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.783391095707812e-06,
+      "loss": 0.1006,
+      "step": 1725
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.783257797920556e-06,
+      "loss": 0.1517,
+      "step": 1726
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.7831245001333e-06,
+      "loss": 0.1064,
+      "step": 1727
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.782991202346042e-06,
+      "loss": 0.1322,
+      "step": 1728
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.782857904558785e-06,
+      "loss": 0.0953,
+      "step": 1729
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.78272460677153e-06,
+      "loss": 0.1349,
+      "step": 1730
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.782591308984272e-06,
+      "loss": 0.1127,
+      "step": 1731
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.782458011197015e-06,
+      "loss": 0.1106,
+      "step": 1732
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.782324713409758e-06,
+      "loss": 0.1498,
+      "step": 1733
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.782191415622502e-06,
+      "loss": 0.1063,
+      "step": 1734
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.782058117835245e-06,
+      "loss": 0.1406,
+      "step": 1735
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.781924820047988e-06,
+      "loss": 0.1521,
+      "step": 1736
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.781791522260732e-06,
+      "loss": 0.1195,
+      "step": 1737
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.781658224473475e-06,
+      "loss": 0.1207,
+      "step": 1738
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.781524926686218e-06,
+      "loss": 0.1391,
+      "step": 1739
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.781391628898961e-06,
+      "loss": 0.1398,
+      "step": 1740
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.781258331111705e-06,
+      "loss": 0.1228,
+      "step": 1741
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.781125033324448e-06,
+      "loss": 0.1363,
+      "step": 1742
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.780991735537191e-06,
+      "loss": 0.1378,
+      "step": 1743
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.780858437749934e-06,
+      "loss": 0.1119,
+      "step": 1744
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.780725139962678e-06,
+      "loss": 0.155,
+      "step": 1745
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.780591842175421e-06,
+      "loss": 0.1424,
+      "step": 1746
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.780458544388164e-06,
+      "loss": 0.1536,
+      "step": 1747
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.780325246600907e-06,
+      "loss": 0.0943,
+      "step": 1748
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.780191948813651e-06,
+      "loss": 0.1506,
+      "step": 1749
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.780058651026394e-06,
+      "loss": 0.1132,
+      "step": 1750
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.779925353239137e-06,
+      "loss": 0.1422,
+      "step": 1751
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.77979205545188e-06,
+      "loss": 0.1338,
+      "step": 1752
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.779658757664624e-06,
+      "loss": 0.1814,
+      "step": 1753
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.779525459877367e-06,
+      "loss": 0.12,
+      "step": 1754
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.77939216209011e-06,
+      "loss": 0.1689,
+      "step": 1755
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.779258864302853e-06,
+      "loss": 0.1059,
+      "step": 1756
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.779125566515597e-06,
+      "loss": 0.1362,
+      "step": 1757
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.77899226872834e-06,
+      "loss": 0.1212,
+      "step": 1758
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.778858970941083e-06,
+      "loss": 0.1392,
+      "step": 1759
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.778725673153826e-06,
+      "loss": 0.1904,
+      "step": 1760
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.77859237536657e-06,
+      "loss": 0.1618,
+      "step": 1761
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.778459077579313e-06,
+      "loss": 0.121,
+      "step": 1762
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.778325779792056e-06,
+      "loss": 0.1183,
+      "step": 1763
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.778192482004799e-06,
+      "loss": 0.138,
+      "step": 1764
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.778059184217543e-06,
+      "loss": 0.1561,
+      "step": 1765
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.777925886430286e-06,
+      "loss": 0.1445,
+      "step": 1766
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.777792588643029e-06,
+      "loss": 0.1366,
+      "step": 1767
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.777659290855772e-06,
+      "loss": 0.1508,
+      "step": 1768
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.777525993068516e-06,
+      "loss": 0.1167,
+      "step": 1769
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.777392695281259e-06,
+      "loss": 0.1271,
+      "step": 1770
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.777259397494002e-06,
+      "loss": 0.139,
+      "step": 1771
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.777126099706745e-06,
+      "loss": 0.1475,
+      "step": 1772
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.77699280191949e-06,
+      "loss": 0.1098,
+      "step": 1773
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.776859504132232e-06,
+      "loss": 0.1701,
+      "step": 1774
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.776726206344975e-06,
+      "loss": 0.1366,
+      "step": 1775
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.776592908557718e-06,
+      "loss": 0.1266,
+      "step": 1776
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.776459610770462e-06,
+      "loss": 0.1086,
+      "step": 1777
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.776326312983205e-06,
+      "loss": 0.1463,
+      "step": 1778
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.776193015195948e-06,
+      "loss": 0.1563,
+      "step": 1779
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.77605971740869e-06,
+      "loss": 0.1593,
+      "step": 1780
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.775926419621435e-06,
+      "loss": 0.1155,
+      "step": 1781
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.775793121834178e-06,
+      "loss": 0.1376,
+      "step": 1782
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.77565982404692e-06,
+      "loss": 0.135,
+      "step": 1783
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.775526526259664e-06,
+      "loss": 0.1665,
+      "step": 1784
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.775393228472408e-06,
+      "loss": 0.111,
+      "step": 1785
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.775259930685151e-06,
+      "loss": 0.1414,
+      "step": 1786
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.775126632897894e-06,
+      "loss": 0.1146,
+      "step": 1787
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.774993335110638e-06,
+      "loss": 0.1146,
+      "step": 1788
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.774860037323381e-06,
+      "loss": 0.1142,
+      "step": 1789
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.774726739536124e-06,
+      "loss": 0.1382,
+      "step": 1790
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.774593441748867e-06,
+      "loss": 0.1291,
+      "step": 1791
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.774460143961611e-06,
+      "loss": 0.1205,
+      "step": 1792
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.774326846174354e-06,
+      "loss": 0.1495,
+      "step": 1793
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.774193548387097e-06,
+      "loss": 0.1579,
+      "step": 1794
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.774060250599841e-06,
+      "loss": 0.1323,
+      "step": 1795
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.773926952812584e-06,
+      "loss": 0.1395,
+      "step": 1796
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.773793655025327e-06,
+      "loss": 0.1179,
+      "step": 1797
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.77366035723807e-06,
+      "loss": 0.1135,
+      "step": 1798
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.773527059450814e-06,
+      "loss": 0.2038,
+      "step": 1799
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.773393761663557e-06,
+      "loss": 0.1902,
+      "step": 1800
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.7732604638763e-06,
+      "loss": 0.119,
+      "step": 1801
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.773127166089044e-06,
+      "loss": 0.1534,
+      "step": 1802
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.772993868301787e-06,
+      "loss": 0.1078,
+      "step": 1803
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.77286057051453e-06,
+      "loss": 0.0912,
+      "step": 1804
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.772727272727273e-06,
+      "loss": 0.1102,
+      "step": 1805
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.772593974940017e-06,
+      "loss": 0.1384,
+      "step": 1806
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.77246067715276e-06,
+      "loss": 0.1403,
+      "step": 1807
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.772327379365503e-06,
+      "loss": 0.1266,
+      "step": 1808
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.772194081578248e-06,
+      "loss": 0.1436,
+      "step": 1809
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.77206078379099e-06,
+      "loss": 0.167,
+      "step": 1810
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.771927486003733e-06,
+      "loss": 0.1505,
+      "step": 1811
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.771794188216476e-06,
+      "loss": 0.1627,
+      "step": 1812
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.77166089042922e-06,
+      "loss": 0.1118,
+      "step": 1813
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.771527592641963e-06,
+      "loss": 0.1544,
+      "step": 1814
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.771394294854706e-06,
+      "loss": 0.1682,
+      "step": 1815
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.77126099706745e-06,
+      "loss": 0.1495,
+      "step": 1816
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.771127699280194e-06,
+      "loss": 0.1757,
+      "step": 1817
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.770994401492936e-06,
+      "loss": 0.1483,
+      "step": 1818
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.77086110370568e-06,
+      "loss": 0.107,
+      "step": 1819
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.770727805918424e-06,
+      "loss": 0.1152,
+      "step": 1820
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.770594508131167e-06,
+      "loss": 0.1475,
+      "step": 1821
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.77046121034391e-06,
+      "loss": 0.1418,
+      "step": 1822
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.770327912556652e-06,
+      "loss": 0.146,
+      "step": 1823
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.770194614769397e-06,
+      "loss": 0.1206,
+      "step": 1824
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.77006131698214e-06,
+      "loss": 0.1792,
+      "step": 1825
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.769928019194882e-06,
+      "loss": 0.1259,
+      "step": 1826
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.769794721407625e-06,
+      "loss": 0.1304,
+      "step": 1827
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.76966142362037e-06,
+      "loss": 0.1385,
+      "step": 1828
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.769528125833112e-06,
+      "loss": 0.1205,
+      "step": 1829
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.769394828045855e-06,
+      "loss": 0.0905,
+      "step": 1830
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.769261530258598e-06,
+      "loss": 0.137,
+      "step": 1831
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.769128232471343e-06,
+      "loss": 0.1113,
+      "step": 1832
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.768994934684085e-06,
+      "loss": 0.0988,
+      "step": 1833
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.768861636896828e-06,
+      "loss": 0.1314,
+      "step": 1834
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.768728339109571e-06,
+      "loss": 0.1535,
+      "step": 1835
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.768595041322316e-06,
+      "loss": 0.1246,
+      "step": 1836
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.768461743535058e-06,
+      "loss": 0.1274,
+      "step": 1837
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.768328445747801e-06,
+      "loss": 0.1472,
+      "step": 1838
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.768195147960544e-06,
+      "loss": 0.1248,
+      "step": 1839
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.768061850173289e-06,
+      "loss": 0.1242,
+      "step": 1840
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.767928552386031e-06,
+      "loss": 0.1134,
+      "step": 1841
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.767795254598774e-06,
+      "loss": 0.1543,
+      "step": 1842
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.767661956811517e-06,
+      "loss": 0.174,
+      "step": 1843
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.767528659024262e-06,
+      "loss": 0.1264,
+      "step": 1844
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.767395361237004e-06,
+      "loss": 0.1095,
+      "step": 1845
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.767262063449747e-06,
+      "loss": 0.1232,
+      "step": 1846
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.767128765662492e-06,
+      "loss": 0.0957,
+      "step": 1847
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.766995467875234e-06,
+      "loss": 0.1329,
+      "step": 1848
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.766862170087977e-06,
+      "loss": 0.0959,
+      "step": 1849
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.76672887230072e-06,
+      "loss": 0.1291,
+      "step": 1850
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.766595574513465e-06,
+      "loss": 0.12,
+      "step": 1851
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.766462276726207e-06,
+      "loss": 0.1454,
+      "step": 1852
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.76632897893895e-06,
+      "loss": 0.1606,
+      "step": 1853
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.766195681151693e-06,
+      "loss": 0.1409,
+      "step": 1854
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.766062383364438e-06,
+      "loss": 0.1063,
+      "step": 1855
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.76592908557718e-06,
+      "loss": 0.1121,
+      "step": 1856
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.765795787789923e-06,
+      "loss": 0.1548,
+      "step": 1857
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.765662490002666e-06,
+      "loss": 0.1909,
+      "step": 1858
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.76552919221541e-06,
+      "loss": 0.115,
+      "step": 1859
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.765395894428153e-06,
+      "loss": 0.1573,
+      "step": 1860
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.765262596640896e-06,
+      "loss": 0.1227,
+      "step": 1861
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.765129298853639e-06,
+      "loss": 0.1872,
+      "step": 1862
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.764996001066384e-06,
+      "loss": 0.1069,
+      "step": 1863
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.764862703279126e-06,
+      "loss": 0.1331,
+      "step": 1864
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.76472940549187e-06,
+      "loss": 0.1046,
+      "step": 1865
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.764596107704612e-06,
+      "loss": 0.1954,
+      "step": 1866
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.764462809917357e-06,
+      "loss": 0.103,
+      "step": 1867
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.7643295121301e-06,
+      "loss": 0.1296,
+      "step": 1868
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.764196214342842e-06,
+      "loss": 0.1528,
+      "step": 1869
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.764062916555585e-06,
+      "loss": 0.1346,
+      "step": 1870
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.76392961876833e-06,
+      "loss": 0.0964,
+      "step": 1871
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.763796320981072e-06,
+      "loss": 0.1312,
+      "step": 1872
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.763663023193815e-06,
+      "loss": 0.1098,
+      "step": 1873
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.763529725406558e-06,
+      "loss": 0.1417,
+      "step": 1874
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.763396427619302e-06,
+      "loss": 0.1186,
+      "step": 1875
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.763263129832045e-06,
+      "loss": 0.1102,
+      "step": 1876
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.763129832044788e-06,
+      "loss": 0.1238,
+      "step": 1877
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.762996534257531e-06,
+      "loss": 0.1405,
+      "step": 1878
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.762863236470275e-06,
+      "loss": 0.1778,
+      "step": 1879
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.762729938683018e-06,
+      "loss": 0.1354,
+      "step": 1880
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.762596640895761e-06,
+      "loss": 0.1317,
+      "step": 1881
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.762463343108506e-06,
+      "loss": 0.1562,
+      "step": 1882
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.762330045321248e-06,
+      "loss": 0.1528,
+      "step": 1883
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.762196747533991e-06,
+      "loss": 0.1447,
+      "step": 1884
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.762063449746734e-06,
+      "loss": 0.1243,
+      "step": 1885
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.761930151959479e-06,
+      "loss": 0.1213,
+      "step": 1886
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.761796854172221e-06,
+      "loss": 0.1456,
+      "step": 1887
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.761663556384964e-06,
+      "loss": 0.1662,
+      "step": 1888
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.761530258597709e-06,
+      "loss": 0.1223,
+      "step": 1889
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.761396960810452e-06,
+      "loss": 0.1864,
+      "step": 1890
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.761263663023194e-06,
+      "loss": 0.1207,
+      "step": 1891
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.761130365235937e-06,
+      "loss": 0.1477,
+      "step": 1892
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.760997067448682e-06,
+      "loss": 0.0921,
+      "step": 1893
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.760863769661424e-06,
+      "loss": 0.1602,
+      "step": 1894
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.760730471874167e-06,
+      "loss": 0.1454,
+      "step": 1895
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.760597174086912e-06,
+      "loss": 0.0986,
+      "step": 1896
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.760463876299655e-06,
+      "loss": 0.1073,
+      "step": 1897
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.760330578512397e-06,
+      "loss": 0.1261,
+      "step": 1898
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.76019728072514e-06,
+      "loss": 0.1233,
+      "step": 1899
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.760063982937885e-06,
+      "loss": 0.123,
+      "step": 1900
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.759930685150628e-06,
+      "loss": 0.1512,
+      "step": 1901
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.75979738736337e-06,
+      "loss": 0.0907,
+      "step": 1902
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.759664089576115e-06,
+      "loss": 0.1465,
+      "step": 1903
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.759530791788858e-06,
+      "loss": 0.0993,
+      "step": 1904
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.7593974940016e-06,
+      "loss": 0.1419,
+      "step": 1905
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.759264196214343e-06,
+      "loss": 0.0921,
+      "step": 1906
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.759130898427088e-06,
+      "loss": 0.1457,
+      "step": 1907
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.75899760063983e-06,
+      "loss": 0.1238,
+      "step": 1908
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.758864302852574e-06,
+      "loss": 0.1355,
+      "step": 1909
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.758731005065318e-06,
+      "loss": 0.1516,
+      "step": 1910
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.758597707278061e-06,
+      "loss": 0.1598,
+      "step": 1911
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.758464409490804e-06,
+      "loss": 0.1406,
+      "step": 1912
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.758331111703547e-06,
+      "loss": 0.1293,
+      "step": 1913
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.758197813916291e-06,
+      "loss": 0.096,
+      "step": 1914
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.758064516129034e-06,
+      "loss": 0.1782,
+      "step": 1915
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.757931218341777e-06,
+      "loss": 0.1232,
+      "step": 1916
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.75779792055452e-06,
+      "loss": 0.1581,
+      "step": 1917
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.757664622767264e-06,
+      "loss": 0.1308,
+      "step": 1918
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.757531324980007e-06,
+      "loss": 0.1211,
+      "step": 1919
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.75739802719275e-06,
+      "loss": 0.1317,
+      "step": 1920
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.757264729405492e-06,
+      "loss": 0.1216,
+      "step": 1921
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.757131431618237e-06,
+      "loss": 0.1807,
+      "step": 1922
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.75699813383098e-06,
+      "loss": 0.1097,
+      "step": 1923
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.756864836043723e-06,
+      "loss": 0.147,
+      "step": 1924
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.756731538256465e-06,
+      "loss": 0.1341,
+      "step": 1925
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.75659824046921e-06,
+      "loss": 0.1203,
+      "step": 1926
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.756464942681953e-06,
+      "loss": 0.1331,
+      "step": 1927
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.756331644894696e-06,
+      "loss": 0.1524,
+      "step": 1928
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.756198347107438e-06,
+      "loss": 0.1224,
+      "step": 1929
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.756065049320183e-06,
+      "loss": 0.1262,
+      "step": 1930
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.755931751532926e-06,
+      "loss": 0.1611,
+      "step": 1931
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.755798453745669e-06,
+      "loss": 0.098,
+      "step": 1932
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.755665155958411e-06,
+      "loss": 0.1169,
+      "step": 1933
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.755531858171156e-06,
+      "loss": 0.1093,
+      "step": 1934
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.755398560383899e-06,
+      "loss": 0.2282,
+      "step": 1935
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.755265262596642e-06,
+      "loss": 0.1635,
+      "step": 1936
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.755131964809384e-06,
+      "loss": 0.1107,
+      "step": 1937
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.754998667022129e-06,
+      "loss": 0.1308,
+      "step": 1938
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.754865369234872e-06,
+      "loss": 0.1069,
+      "step": 1939
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.754732071447615e-06,
+      "loss": 0.1555,
+      "step": 1940
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.754598773660357e-06,
+      "loss": 0.1802,
+      "step": 1941
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.754465475873102e-06,
+      "loss": 0.1327,
+      "step": 1942
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.754332178085845e-06,
+      "loss": 0.1564,
+      "step": 1943
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.754198880298587e-06,
+      "loss": 0.1397,
+      "step": 1944
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.75406558251133e-06,
+      "loss": 0.1214,
+      "step": 1945
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.753932284724075e-06,
+      "loss": 0.1168,
+      "step": 1946
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.753798986936818e-06,
+      "loss": 0.1174,
+      "step": 1947
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.75366568914956e-06,
+      "loss": 0.102,
+      "step": 1948
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.753532391362303e-06,
+      "loss": 0.1305,
+      "step": 1949
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.753399093575048e-06,
+      "loss": 0.1607,
+      "step": 1950
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.75326579578779e-06,
+      "loss": 0.1181,
+      "step": 1951
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.753132498000533e-06,
+      "loss": 0.1243,
+      "step": 1952
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.752999200213276e-06,
+      "loss": 0.1055,
+      "step": 1953
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.75286590242602e-06,
+      "loss": 0.0914,
+      "step": 1954
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.752732604638764e-06,
+      "loss": 0.1298,
+      "step": 1955
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.752599306851506e-06,
+      "loss": 0.1078,
+      "step": 1956
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.75246600906425e-06,
+      "loss": 0.1526,
+      "step": 1957
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.752332711276994e-06,
+      "loss": 0.126,
+      "step": 1958
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.752199413489737e-06,
+      "loss": 0.1312,
+      "step": 1959
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.75206611570248e-06,
+      "loss": 0.1199,
+      "step": 1960
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.751932817915224e-06,
+      "loss": 0.1334,
+      "step": 1961
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.751799520127967e-06,
+      "loss": 0.1161,
+      "step": 1962
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.75166622234071e-06,
+      "loss": 0.1195,
+      "step": 1963
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.751532924553452e-06,
+      "loss": 0.1148,
+      "step": 1964
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.751399626766197e-06,
+      "loss": 0.1487,
+      "step": 1965
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.75126632897894e-06,
+      "loss": 0.0934,
+      "step": 1966
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.751133031191682e-06,
+      "loss": 0.1215,
+      "step": 1967
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.750999733404425e-06,
+      "loss": 0.1432,
+      "step": 1968
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.75086643561717e-06,
+      "loss": 0.1575,
+      "step": 1969
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.750733137829913e-06,
+      "loss": 0.1273,
+      "step": 1970
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.750599840042655e-06,
+      "loss": 0.1115,
+      "step": 1971
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.750466542255398e-06,
+      "loss": 0.1316,
+      "step": 1972
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.750333244468143e-06,
+      "loss": 0.1476,
+      "step": 1973
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.750199946680886e-06,
+      "loss": 0.1389,
+      "step": 1974
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.750066648893628e-06,
+      "loss": 0.1126,
+      "step": 1975
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.749933351106371e-06,
+      "loss": 0.1639,
+      "step": 1976
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.749800053319116e-06,
+      "loss": 0.1397,
+      "step": 1977
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.749666755531859e-06,
+      "loss": 0.1299,
+      "step": 1978
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.749533457744601e-06,
+      "loss": 0.1223,
+      "step": 1979
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.749400159957346e-06,
+      "loss": 0.1389,
+      "step": 1980
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.749266862170089e-06,
+      "loss": 0.1702,
+      "step": 1981
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.749133564382832e-06,
+      "loss": 0.118,
+      "step": 1982
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.749000266595574e-06,
+      "loss": 0.152,
+      "step": 1983
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.748866968808319e-06,
+      "loss": 0.145,
+      "step": 1984
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.748733671021062e-06,
+      "loss": 0.1117,
+      "step": 1985
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.748600373233805e-06,
+      "loss": 0.1094,
+      "step": 1986
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.748467075446549e-06,
+      "loss": 0.1378,
+      "step": 1987
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.748333777659292e-06,
+      "loss": 0.1125,
+      "step": 1988
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.748200479872035e-06,
+      "loss": 0.1638,
+      "step": 1989
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.748067182084777e-06,
+      "loss": 0.182,
+      "step": 1990
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.747933884297522e-06,
+      "loss": 0.1632,
+      "step": 1991
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.747800586510265e-06,
+      "loss": 0.1147,
+      "step": 1992
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.747667288723008e-06,
+      "loss": 0.1459,
+      "step": 1993
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.747533990935752e-06,
+      "loss": 0.1286,
+      "step": 1994
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.747400693148495e-06,
+      "loss": 0.1447,
+      "step": 1995
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.747267395361238e-06,
+      "loss": 0.1019,
+      "step": 1996
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.74713409757398e-06,
+      "loss": 0.1079,
+      "step": 1997
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.747000799786725e-06,
+      "loss": 0.1023,
+      "step": 1998
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.746867501999468e-06,
+      "loss": 0.1252,
+      "step": 1999
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.74673420421221e-06,
+      "loss": 0.1265,
+      "step": 2000
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.746600906424955e-06,
+      "loss": 0.152,
+      "step": 2001
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.746467608637698e-06,
+      "loss": 0.1737,
+      "step": 2002
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.746334310850441e-06,
+      "loss": 0.1302,
+      "step": 2003
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.746201013063184e-06,
+      "loss": 0.1398,
+      "step": 2004
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.746067715275928e-06,
+      "loss": 0.142,
+      "step": 2005
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.745934417488671e-06,
+      "loss": 0.1437,
+      "step": 2006
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.745801119701414e-06,
+      "loss": 0.1283,
+      "step": 2007
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.745667821914157e-06,
+      "loss": 0.1607,
+      "step": 2008
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.745534524126901e-06,
+      "loss": 0.1137,
+      "step": 2009
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.745401226339644e-06,
+      "loss": 0.1553,
+      "step": 2010
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.745267928552387e-06,
+      "loss": 0.1159,
+      "step": 2011
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.74513463076513e-06,
+      "loss": 0.092,
+      "step": 2012
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.745001332977874e-06,
+      "loss": 0.1683,
+      "step": 2013
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.744868035190617e-06,
+      "loss": 0.1111,
+      "step": 2014
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.74473473740336e-06,
+      "loss": 0.1324,
+      "step": 2015
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.744601439616104e-06,
+      "loss": 0.109,
+      "step": 2016
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.744468141828847e-06,
+      "loss": 0.1253,
+      "step": 2017
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.74433484404159e-06,
+      "loss": 0.111,
+      "step": 2018
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.744201546254333e-06,
+      "loss": 0.1493,
+      "step": 2019
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.744068248467077e-06,
+      "loss": 0.1208,
+      "step": 2020
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.74393495067982e-06,
+      "loss": 0.1318,
+      "step": 2021
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.743801652892563e-06,
+      "loss": 0.1168,
+      "step": 2022
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.743668355105306e-06,
+      "loss": 0.0886,
+      "step": 2023
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.74353505731805e-06,
+      "loss": 0.1234,
+      "step": 2024
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.743401759530793e-06,
+      "loss": 0.1899,
+      "step": 2025
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.743268461743536e-06,
+      "loss": 0.1265,
+      "step": 2026
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.743135163956279e-06,
+      "loss": 0.1014,
+      "step": 2027
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.743001866169023e-06,
+      "loss": 0.1042,
+      "step": 2028
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.742868568381766e-06,
+      "loss": 0.1498,
+      "step": 2029
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.742735270594509e-06,
+      "loss": 0.1261,
+      "step": 2030
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.742601972807252e-06,
+      "loss": 0.1411,
+      "step": 2031
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.742468675019996e-06,
+      "loss": 0.151,
+      "step": 2032
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.742335377232739e-06,
+      "loss": 0.1392,
+      "step": 2033
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.742202079445482e-06,
+      "loss": 0.1163,
+      "step": 2034
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.742068781658225e-06,
+      "loss": 0.1659,
+      "step": 2035
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.74193548387097e-06,
+      "loss": 0.2141,
+      "step": 2036
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.741802186083712e-06,
+      "loss": 0.1776,
+      "step": 2037
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.741668888296455e-06,
+      "loss": 0.1016,
+      "step": 2038
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.741535590509198e-06,
+      "loss": 0.0897,
+      "step": 2039
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.741402292721942e-06,
+      "loss": 0.1122,
+      "step": 2040
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.741268994934685e-06,
+      "loss": 0.1201,
+      "step": 2041
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.741135697147428e-06,
+      "loss": 0.1289,
+      "step": 2042
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.74100239936017e-06,
+      "loss": 0.1059,
+      "step": 2043
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.740869101572915e-06,
+      "loss": 0.0845,
+      "step": 2044
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.740735803785658e-06,
+      "loss": 0.1439,
+      "step": 2045
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.7406025059984e-06,
+      "loss": 0.136,
+      "step": 2046
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.740469208211144e-06,
+      "loss": 0.1126,
+      "step": 2047
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.740335910423888e-06,
+      "loss": 0.1516,
+      "step": 2048
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.740202612636631e-06,
+      "loss": 0.1301,
+      "step": 2049
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.740069314849374e-06,
+      "loss": 0.1605,
+      "step": 2050
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.739936017062117e-06,
+      "loss": 0.1062,
+      "step": 2051
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.739802719274861e-06,
+      "loss": 0.1162,
+      "step": 2052
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.739669421487604e-06,
+      "loss": 0.1009,
+      "step": 2053
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.739536123700347e-06,
+      "loss": 0.1483,
+      "step": 2054
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.73940282591309e-06,
+      "loss": 0.0846,
+      "step": 2055
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.739269528125834e-06,
+      "loss": 0.1309,
+      "step": 2056
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.739136230338577e-06,
+      "loss": 0.0959,
+      "step": 2057
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.73900293255132e-06,
+      "loss": 0.1103,
+      "step": 2058
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.738869634764062e-06,
+      "loss": 0.1448,
+      "step": 2059
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.738736336976807e-06,
+      "loss": 0.0846,
+      "step": 2060
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.73860303918955e-06,
+      "loss": 0.1239,
+      "step": 2061
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.738469741402293e-06,
+      "loss": 0.1405,
+      "step": 2062
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.738336443615035e-06,
+      "loss": 0.1084,
+      "step": 2063
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.73820314582778e-06,
+      "loss": 0.1286,
+      "step": 2064
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.738069848040523e-06,
+      "loss": 0.1372,
+      "step": 2065
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.737936550253266e-06,
+      "loss": 0.0977,
+      "step": 2066
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.73780325246601e-06,
+      "loss": 0.1504,
+      "step": 2067
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.737669954678753e-06,
+      "loss": 0.0934,
+      "step": 2068
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.737536656891496e-06,
+      "loss": 0.1323,
+      "step": 2069
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.737403359104239e-06,
+      "loss": 0.1554,
+      "step": 2070
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.737270061316983e-06,
+      "loss": 0.1157,
+      "step": 2071
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.737136763529726e-06,
+      "loss": 0.1569,
+      "step": 2072
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.737003465742469e-06,
+      "loss": 0.1217,
+      "step": 2073
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.736870167955213e-06,
+      "loss": 0.1317,
+      "step": 2074
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.736736870167956e-06,
+      "loss": 0.126,
+      "step": 2075
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.736603572380699e-06,
+      "loss": 0.1192,
+      "step": 2076
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.736470274593442e-06,
+      "loss": 0.1483,
+      "step": 2077
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.736336976806186e-06,
+      "loss": 0.1427,
+      "step": 2078
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.736203679018929e-06,
+      "loss": 0.1232,
+      "step": 2079
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.736070381231672e-06,
+      "loss": 0.1311,
+      "step": 2080
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.735937083444416e-06,
+      "loss": 0.1222,
+      "step": 2081
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.73580378565716e-06,
+      "loss": 0.1212,
+      "step": 2082
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.735670487869902e-06,
+      "loss": 0.1038,
+      "step": 2083
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.735537190082645e-06,
+      "loss": 0.1458,
+      "step": 2084
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.73540389229539e-06,
+      "loss": 0.1092,
+      "step": 2085
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.735270594508132e-06,
+      "loss": 0.1262,
+      "step": 2086
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.735137296720875e-06,
+      "loss": 0.1153,
+      "step": 2087
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.73500399893362e-06,
+      "loss": 0.1174,
+      "step": 2088
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.734870701146362e-06,
+      "loss": 0.1031,
+      "step": 2089
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.734737403359105e-06,
+      "loss": 0.0967,
+      "step": 2090
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.734604105571848e-06,
+      "loss": 0.1313,
+      "step": 2091
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.734470807784592e-06,
+      "loss": 0.1373,
+      "step": 2092
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.734337509997335e-06,
+      "loss": 0.1115,
+      "step": 2093
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.734204212210078e-06,
+      "loss": 0.1507,
+      "step": 2094
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.734070914422823e-06,
+      "loss": 0.1505,
+      "step": 2095
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.733937616635565e-06,
+      "loss": 0.1243,
+      "step": 2096
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.733804318848308e-06,
+      "loss": 0.1649,
+      "step": 2097
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.733671021061051e-06,
+      "loss": 0.124,
+      "step": 2098
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.733537723273796e-06,
+      "loss": 0.125,
+      "step": 2099
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.733404425486538e-06,
+      "loss": 0.138,
+      "step": 2100
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.733271127699281e-06,
+      "loss": 0.1014,
+      "step": 2101
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.733137829912024e-06,
+      "loss": 0.1536,
+      "step": 2102
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.733004532124769e-06,
+      "loss": 0.1865,
+      "step": 2103
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.732871234337511e-06,
+      "loss": 0.1425,
+      "step": 2104
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.732737936550254e-06,
+      "loss": 0.1651,
+      "step": 2105
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.732604638762997e-06,
+      "loss": 0.1046,
+      "step": 2106
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.732471340975742e-06,
+      "loss": 0.1701,
+      "step": 2107
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.732338043188484e-06,
+      "loss": 0.1356,
+      "step": 2108
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.732204745401227e-06,
+      "loss": 0.1451,
+      "step": 2109
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.73207144761397e-06,
+      "loss": 0.1216,
+      "step": 2110
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.731938149826714e-06,
+      "loss": 0.1562,
+      "step": 2111
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.731804852039457e-06,
+      "loss": 0.1224,
+      "step": 2112
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.7316715542522e-06,
+      "loss": 0.0921,
+      "step": 2113
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.731538256464943e-06,
+      "loss": 0.1142,
+      "step": 2114
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.731404958677687e-06,
+      "loss": 0.1319,
+      "step": 2115
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.73127166089043e-06,
+      "loss": 0.1533,
+      "step": 2116
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.731138363103173e-06,
+      "loss": 0.1773,
+      "step": 2117
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.731005065315916e-06,
+      "loss": 0.1268,
+      "step": 2118
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.73087176752866e-06,
+      "loss": 0.1233,
+      "step": 2119
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.730738469741403e-06,
+      "loss": 0.1082,
+      "step": 2120
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.730605171954146e-06,
+      "loss": 0.1402,
+      "step": 2121
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.730471874166889e-06,
+      "loss": 0.1084,
+      "step": 2122
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.730338576379633e-06,
+      "loss": 0.1352,
+      "step": 2123
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.730205278592376e-06,
+      "loss": 0.1464,
+      "step": 2124
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.730071980805119e-06,
+      "loss": 0.1202,
+      "step": 2125
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.729938683017862e-06,
+      "loss": 0.1292,
+      "step": 2126
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.729805385230606e-06,
+      "loss": 0.1113,
+      "step": 2127
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.72967208744335e-06,
+      "loss": 0.1326,
+      "step": 2128
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.729538789656092e-06,
+      "loss": 0.0887,
+      "step": 2129
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.729405491868837e-06,
+      "loss": 0.145,
+      "step": 2130
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.72927219408158e-06,
+      "loss": 0.1211,
+      "step": 2131
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.729138896294322e-06,
+      "loss": 0.1509,
+      "step": 2132
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.729005598507065e-06,
+      "loss": 0.133,
+      "step": 2133
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.72887230071981e-06,
+      "loss": 0.0972,
+      "step": 2134
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.728739002932552e-06,
+      "loss": 0.1394,
+      "step": 2135
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.728605705145295e-06,
+      "loss": 0.1622,
+      "step": 2136
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.728472407358038e-06,
+      "loss": 0.1159,
+      "step": 2137
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.728339109570782e-06,
+      "loss": 0.1404,
+      "step": 2138
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.728205811783525e-06,
+      "loss": 0.0883,
+      "step": 2139
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.728072513996268e-06,
+      "loss": 0.1164,
+      "step": 2140
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.727939216209011e-06,
+      "loss": 0.12,
+      "step": 2141
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.727805918421755e-06,
+      "loss": 0.1356,
+      "step": 2142
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.727672620634498e-06,
+      "loss": 0.1226,
+      "step": 2143
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.727539322847241e-06,
+      "loss": 0.2081,
+      "step": 2144
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.727406025059984e-06,
+      "loss": 0.1468,
+      "step": 2145
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.727272727272728e-06,
+      "loss": 0.1634,
+      "step": 2146
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.727139429485471e-06,
+      "loss": 0.135,
+      "step": 2147
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.727006131698214e-06,
+      "loss": 0.1445,
+      "step": 2148
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.726872833910957e-06,
+      "loss": 0.1411,
+      "step": 2149
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.726739536123701e-06,
+      "loss": 0.1244,
+      "step": 2150
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.726606238336444e-06,
+      "loss": 0.1516,
+      "step": 2151
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.726472940549187e-06,
+      "loss": 0.137,
+      "step": 2152
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.72633964276193e-06,
+      "loss": 0.0863,
+      "step": 2153
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.726206344974674e-06,
+      "loss": 0.0895,
+      "step": 2154
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.726073047187417e-06,
+      "loss": 0.135,
+      "step": 2155
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.72593974940016e-06,
+      "loss": 0.1499,
+      "step": 2156
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.725806451612903e-06,
+      "loss": 0.125,
+      "step": 2157
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.725673153825647e-06,
+      "loss": 0.1397,
+      "step": 2158
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.72553985603839e-06,
+      "loss": 0.1215,
+      "step": 2159
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.725406558251133e-06,
+      "loss": 0.117,
+      "step": 2160
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.725273260463877e-06,
+      "loss": 0.1159,
+      "step": 2161
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.72513996267662e-06,
+      "loss": 0.1108,
+      "step": 2162
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.725006664889363e-06,
+      "loss": 0.1352,
+      "step": 2163
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.724873367102106e-06,
+      "loss": 0.1298,
+      "step": 2164
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.72474006931485e-06,
+      "loss": 0.0934,
+      "step": 2165
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.724606771527593e-06,
+      "loss": 0.1384,
+      "step": 2166
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.724473473740336e-06,
+      "loss": 0.1015,
+      "step": 2167
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.72434017595308e-06,
+      "loss": 0.122,
+      "step": 2168
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.724206878165823e-06,
+      "loss": 0.1132,
+      "step": 2169
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.724073580378566e-06,
+      "loss": 0.1103,
+      "step": 2170
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.723940282591309e-06,
+      "loss": 0.1221,
+      "step": 2171
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.723806984804054e-06,
+      "loss": 0.129,
+      "step": 2172
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.723673687016796e-06,
+      "loss": 0.095,
+      "step": 2173
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.72354038922954e-06,
+      "loss": 0.1196,
+      "step": 2174
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.723407091442284e-06,
+      "loss": 0.1518,
+      "step": 2175
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.723273793655027e-06,
+      "loss": 0.1067,
+      "step": 2176
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.72314049586777e-06,
+      "loss": 0.1528,
+      "step": 2177
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.723007198080512e-06,
+      "loss": 0.119,
+      "step": 2178
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.722873900293257e-06,
+      "loss": 0.1026,
+      "step": 2179
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.722740602506e-06,
+      "loss": 0.1365,
+      "step": 2180
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.722607304718742e-06,
+      "loss": 0.146,
+      "step": 2181
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.722474006931487e-06,
+      "loss": 0.1439,
+      "step": 2182
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.72234070914423e-06,
+      "loss": 0.0935,
+      "step": 2183
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.722207411356972e-06,
+      "loss": 0.113,
+      "step": 2184
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.722074113569715e-06,
+      "loss": 0.1226,
+      "step": 2185
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.72194081578246e-06,
+      "loss": 0.1286,
+      "step": 2186
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.721807517995203e-06,
+      "loss": 0.1043,
+      "step": 2187
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.721674220207945e-06,
+      "loss": 0.1557,
+      "step": 2188
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.72154092242069e-06,
+      "loss": 0.1515,
+      "step": 2189
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.721407624633433e-06,
+      "loss": 0.1204,
+      "step": 2190
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.721274326846176e-06,
+      "loss": 0.0865,
+      "step": 2191
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.721141029058918e-06,
+      "loss": 0.1234,
+      "step": 2192
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.721007731271663e-06,
+      "loss": 0.0997,
+      "step": 2193
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.720874433484406e-06,
+      "loss": 0.1075,
+      "step": 2194
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.720741135697149e-06,
+      "loss": 0.1163,
+      "step": 2195
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.720607837909891e-06,
+      "loss": 0.1645,
+      "step": 2196
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.720474540122636e-06,
+      "loss": 0.1265,
+      "step": 2197
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.720341242335379e-06,
+      "loss": 0.13,
+      "step": 2198
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.720207944548122e-06,
+      "loss": 0.1119,
+      "step": 2199
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.720074646760864e-06,
+      "loss": 0.1338,
+      "step": 2200
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.719941348973609e-06,
+      "loss": 0.1236,
+      "step": 2201
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.719808051186352e-06,
+      "loss": 0.1225,
+      "step": 2202
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.719674753399094e-06,
+      "loss": 0.128,
+      "step": 2203
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.719541455611837e-06,
+      "loss": 0.0845,
+      "step": 2204
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.719408157824582e-06,
+      "loss": 0.1166,
+      "step": 2205
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.719274860037325e-06,
+      "loss": 0.1787,
+      "step": 2206
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.719141562250067e-06,
+      "loss": 0.1141,
+      "step": 2207
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.71900826446281e-06,
+      "loss": 0.1653,
+      "step": 2208
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.718874966675555e-06,
+      "loss": 0.1087,
+      "step": 2209
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.718741668888298e-06,
+      "loss": 0.1391,
+      "step": 2210
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.71860837110104e-06,
+      "loss": 0.161,
+      "step": 2211
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.718475073313783e-06,
+      "loss": 0.1224,
+      "step": 2212
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.718341775526528e-06,
+      "loss": 0.1015,
+      "step": 2213
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.71820847773927e-06,
+      "loss": 0.1009,
+      "step": 2214
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.718075179952013e-06,
+      "loss": 0.0886,
+      "step": 2215
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.717941882164756e-06,
+      "loss": 0.1379,
+      "step": 2216
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.7178085843775e-06,
+      "loss": 0.1317,
+      "step": 2217
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.717675286590244e-06,
+      "loss": 0.1152,
+      "step": 2218
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.717541988802986e-06,
+      "loss": 0.1021,
+      "step": 2219
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.71740869101573e-06,
+      "loss": 0.1932,
+      "step": 2220
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.717275393228474e-06,
+      "loss": 0.1122,
+      "step": 2221
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.717142095441217e-06,
+      "loss": 0.1156,
+      "step": 2222
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.71700879765396e-06,
+      "loss": 0.1641,
+      "step": 2223
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.716875499866702e-06,
+      "loss": 0.0979,
+      "step": 2224
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.716742202079447e-06,
+      "loss": 0.1247,
+      "step": 2225
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.71660890429219e-06,
+      "loss": 0.1125,
+      "step": 2226
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.716475606504932e-06,
+      "loss": 0.1129,
+      "step": 2227
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.716342308717675e-06,
+      "loss": 0.1222,
+      "step": 2228
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.71620901093042e-06,
+      "loss": 0.1329,
+      "step": 2229
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.716075713143162e-06,
+      "loss": 0.1195,
+      "step": 2230
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.715942415355905e-06,
+      "loss": 0.102,
+      "step": 2231
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.715809117568648e-06,
+      "loss": 0.1405,
+      "step": 2232
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.715675819781393e-06,
+      "loss": 0.178,
+      "step": 2233
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.715542521994135e-06,
+      "loss": 0.1061,
+      "step": 2234
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.715409224206878e-06,
+      "loss": 0.1514,
+      "step": 2235
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.715275926419621e-06,
+      "loss": 0.11,
+      "step": 2236
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.715142628632366e-06,
+      "loss": 0.1555,
+      "step": 2237
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.715009330845108e-06,
+      "loss": 0.1583,
+      "step": 2238
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.714876033057851e-06,
+      "loss": 0.1616,
+      "step": 2239
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.714742735270594e-06,
+      "loss": 0.1238,
+      "step": 2240
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.714609437483339e-06,
+      "loss": 0.1404,
+      "step": 2241
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.714476139696081e-06,
+      "loss": 0.1305,
+      "step": 2242
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.714342841908824e-06,
+      "loss": 0.1051,
+      "step": 2243
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.714209544121569e-06,
+      "loss": 0.0934,
+      "step": 2244
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.714076246334312e-06,
+      "loss": 0.1325,
+      "step": 2245
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.713942948547054e-06,
+      "loss": 0.1249,
+      "step": 2246
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.713809650759797e-06,
+      "loss": 0.1253,
+      "step": 2247
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.713676352972542e-06,
+      "loss": 0.1091,
+      "step": 2248
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.713543055185284e-06,
+      "loss": 0.1365,
+      "step": 2249
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.713409757398027e-06,
+      "loss": 0.1682,
+      "step": 2250
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.71327645961077e-06,
+      "loss": 0.1163,
+      "step": 2251
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.713143161823515e-06,
+      "loss": 0.1492,
+      "step": 2252
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.713009864036257e-06,
+      "loss": 0.1192,
+      "step": 2253
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.712876566249e-06,
+      "loss": 0.127,
+      "step": 2254
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.712743268461745e-06,
+      "loss": 0.1173,
+      "step": 2255
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.712609970674488e-06,
+      "loss": 0.1577,
+      "step": 2256
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.71247667288723e-06,
+      "loss": 0.1141,
+      "step": 2257
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.712343375099973e-06,
+      "loss": 0.122,
+      "step": 2258
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.712210077312718e-06,
+      "loss": 0.1325,
+      "step": 2259
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.71207677952546e-06,
+      "loss": 0.265,
+      "step": 2260
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.711943481738203e-06,
+      "loss": 0.1778,
+      "step": 2261
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.711810183950948e-06,
+      "loss": 0.1511,
+      "step": 2262
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.71167688616369e-06,
+      "loss": 0.1194,
+      "step": 2263
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.711543588376434e-06,
+      "loss": 0.1033,
+      "step": 2264
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.711410290589176e-06,
+      "loss": 0.1107,
+      "step": 2265
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.711276992801921e-06,
+      "loss": 0.1545,
+      "step": 2266
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.711143695014664e-06,
+      "loss": 0.1084,
+      "step": 2267
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.711010397227407e-06,
+      "loss": 0.1276,
+      "step": 2268
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.710877099440151e-06,
+      "loss": 0.1136,
+      "step": 2269
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.710743801652894e-06,
+      "loss": 0.1053,
+      "step": 2270
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.710610503865637e-06,
+      "loss": 0.119,
+      "step": 2271
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.71047720607838e-06,
+      "loss": 0.1589,
+      "step": 2272
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.710343908291124e-06,
+      "loss": 0.1136,
+      "step": 2273
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.710210610503867e-06,
+      "loss": 0.1477,
+      "step": 2274
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.71007731271661e-06,
+      "loss": 0.0959,
+      "step": 2275
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.709944014929354e-06,
+      "loss": 0.1073,
+      "step": 2276
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.709810717142097e-06,
+      "loss": 0.1137,
+      "step": 2277
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.70967741935484e-06,
+      "loss": 0.117,
+      "step": 2278
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.709544121567583e-06,
+      "loss": 0.1249,
+      "step": 2279
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.709410823780327e-06,
+      "loss": 0.1814,
+      "step": 2280
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.70927752599307e-06,
+      "loss": 0.1244,
+      "step": 2281
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.709144228205813e-06,
+      "loss": 0.1344,
+      "step": 2282
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.709010930418556e-06,
+      "loss": 0.1335,
+      "step": 2283
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.7088776326313e-06,
+      "loss": 0.1338,
+      "step": 2284
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.708744334844043e-06,
+      "loss": 0.0972,
+      "step": 2285
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.708611037056786e-06,
+      "loss": 0.1444,
+      "step": 2286
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.708477739269529e-06,
+      "loss": 0.1246,
+      "step": 2287
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.708344441482273e-06,
+      "loss": 0.1062,
+      "step": 2288
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.708211143695016e-06,
+      "loss": 0.1209,
+      "step": 2289
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.708077845907759e-06,
+      "loss": 0.1677,
+      "step": 2290
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.707944548120502e-06,
+      "loss": 0.127,
+      "step": 2291
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.707811250333246e-06,
+      "loss": 0.1708,
+      "step": 2292
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.707677952545989e-06,
+      "loss": 0.1468,
+      "step": 2293
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.707544654758732e-06,
+      "loss": 0.1246,
+      "step": 2294
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.707411356971474e-06,
+      "loss": 0.0963,
+      "step": 2295
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.707278059184219e-06,
+      "loss": 0.1028,
+      "step": 2296
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.707144761396962e-06,
+      "loss": 0.1691,
+      "step": 2297
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.707011463609705e-06,
+      "loss": 0.1136,
+      "step": 2298
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.70687816582245e-06,
+      "loss": 0.1161,
+      "step": 2299
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.706744868035192e-06,
+      "loss": 0.1248,
+      "step": 2300
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.706611570247935e-06,
+      "loss": 0.149,
+      "step": 2301
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.706478272460678e-06,
+      "loss": 0.1277,
+      "step": 2302
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.706344974673422e-06,
+      "loss": 0.1427,
+      "step": 2303
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.706211676886165e-06,
+      "loss": 0.1808,
+      "step": 2304
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.706078379098908e-06,
+      "loss": 0.1468,
+      "step": 2305
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.70594508131165e-06,
+      "loss": 0.1415,
+      "step": 2306
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.705811783524395e-06,
+      "loss": 0.1225,
+      "step": 2307
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.705678485737138e-06,
+      "loss": 0.1066,
+      "step": 2308
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.70554518794988e-06,
+      "loss": 0.0824,
+      "step": 2309
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.705411890162624e-06,
+      "loss": 0.1344,
+      "step": 2310
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.705278592375368e-06,
+      "loss": 0.1665,
+      "step": 2311
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.705145294588111e-06,
+      "loss": 0.1205,
+      "step": 2312
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.705011996800854e-06,
+      "loss": 0.1141,
+      "step": 2313
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.704878699013597e-06,
+      "loss": 0.1145,
+      "step": 2314
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.704745401226341e-06,
+      "loss": 0.1317,
+      "step": 2315
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.704612103439084e-06,
+      "loss": 0.1173,
+      "step": 2316
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.704478805651827e-06,
+      "loss": 0.1274,
+      "step": 2317
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.70434550786457e-06,
+      "loss": 0.091,
+      "step": 2318
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.704212210077314e-06,
+      "loss": 0.1195,
+      "step": 2319
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.704078912290057e-06,
+      "loss": 0.1015,
+      "step": 2320
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.7039456145028e-06,
+      "loss": 0.1615,
+      "step": 2321
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.703812316715542e-06,
+      "loss": 0.1357,
+      "step": 2322
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.703679018928287e-06,
+      "loss": 0.121,
+      "step": 2323
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.70354572114103e-06,
+      "loss": 0.1574,
+      "step": 2324
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.703412423353773e-06,
+      "loss": 0.1341,
+      "step": 2325
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.703279125566515e-06,
+      "loss": 0.1087,
+      "step": 2326
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.70314582777926e-06,
+      "loss": 0.1087,
+      "step": 2327
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.703012529992003e-06,
+      "loss": 0.1257,
+      "step": 2328
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.702879232204746e-06,
+      "loss": 0.1246,
+      "step": 2329
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.702745934417488e-06,
+      "loss": 0.124,
+      "step": 2330
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.702612636630233e-06,
+      "loss": 0.1435,
+      "step": 2331
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.702479338842976e-06,
+      "loss": 0.1345,
+      "step": 2332
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.702346041055719e-06,
+      "loss": 0.1726,
+      "step": 2333
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.702212743268461e-06,
+      "loss": 0.0985,
+      "step": 2334
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.702079445481206e-06,
+      "loss": 0.1115,
+      "step": 2335
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.701946147693949e-06,
+      "loss": 0.1226,
+      "step": 2336
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.701812849906692e-06,
+      "loss": 0.1175,
+      "step": 2337
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.701679552119434e-06,
+      "loss": 0.1271,
+      "step": 2338
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.701546254332179e-06,
+      "loss": 0.1261,
+      "step": 2339
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.701412956544922e-06,
+      "loss": 0.1134,
+      "step": 2340
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.701279658757664e-06,
+      "loss": 0.1401,
+      "step": 2341
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.701146360970409e-06,
+      "loss": 0.1013,
+      "step": 2342
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.701013063183152e-06,
+      "loss": 0.1345,
+      "step": 2343
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.700879765395895e-06,
+      "loss": 0.1216,
+      "step": 2344
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.700746467608637e-06,
+      "loss": 0.1348,
+      "step": 2345
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.700613169821382e-06,
+      "loss": 0.1103,
+      "step": 2346
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.700479872034125e-06,
+      "loss": 0.1227,
+      "step": 2347
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.700346574246868e-06,
+      "loss": 0.132,
+      "step": 2348
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.700213276459612e-06,
+      "loss": 0.1242,
+      "step": 2349
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.700079978672355e-06,
+      "loss": 0.0911,
+      "step": 2350
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.699946680885098e-06,
+      "loss": 0.1149,
+      "step": 2351
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.69981338309784e-06,
+      "loss": 0.0975,
+      "step": 2352
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.699680085310585e-06,
+      "loss": 0.1285,
+      "step": 2353
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.699546787523328e-06,
+      "loss": 0.1599,
+      "step": 2354
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.69941348973607e-06,
+      "loss": 0.1588,
+      "step": 2355
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.699280191948815e-06,
+      "loss": 0.1119,
+      "step": 2356
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.699146894161558e-06,
+      "loss": 0.1386,
+      "step": 2357
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.699013596374301e-06,
+      "loss": 0.1824,
+      "step": 2358
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.698880298587044e-06,
+      "loss": 0.1029,
+      "step": 2359
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.698747000799788e-06,
+      "loss": 0.1512,
+      "step": 2360
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.698613703012531e-06,
+      "loss": 0.1156,
+      "step": 2361
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.698480405225274e-06,
+      "loss": 0.1169,
+      "step": 2362
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.698347107438018e-06,
+      "loss": 0.1595,
+      "step": 2363
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.698213809650761e-06,
+      "loss": 0.0976,
+      "step": 2364
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.698080511863504e-06,
+      "loss": 0.1426,
+      "step": 2365
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.697947214076247e-06,
+      "loss": 0.0931,
+      "step": 2366
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.697813916288991e-06,
+      "loss": 0.1368,
+      "step": 2367
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.697680618501734e-06,
+      "loss": 0.1148,
+      "step": 2368
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.697547320714477e-06,
+      "loss": 0.1021,
+      "step": 2369
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.697414022927221e-06,
+      "loss": 0.1649,
+      "step": 2370
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.697280725139964e-06,
+      "loss": 0.1333,
+      "step": 2371
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.697147427352707e-06,
+      "loss": 0.1124,
+      "step": 2372
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.69701412956545e-06,
+      "loss": 0.1162,
+      "step": 2373
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.696880831778194e-06,
+      "loss": 0.1002,
+      "step": 2374
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.696747533990937e-06,
+      "loss": 0.1342,
+      "step": 2375
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.69661423620368e-06,
+      "loss": 0.1111,
+      "step": 2376
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.696480938416423e-06,
+      "loss": 0.1215,
+      "step": 2377
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.696347640629167e-06,
+      "loss": 0.1777,
+      "step": 2378
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.69621434284191e-06,
+      "loss": 0.1532,
+      "step": 2379
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.696081045054653e-06,
+      "loss": 0.1615,
+      "step": 2380
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.695947747267396e-06,
+      "loss": 0.0897,
+      "step": 2381
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.69581444948014e-06,
+      "loss": 0.1297,
+      "step": 2382
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.695681151692883e-06,
+      "loss": 0.1441,
+      "step": 2383
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.695547853905626e-06,
+      "loss": 0.1142,
+      "step": 2384
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.695414556118369e-06,
+      "loss": 0.0901,
+      "step": 2385
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.695281258331113e-06,
+      "loss": 0.1396,
+      "step": 2386
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.695147960543856e-06,
+      "loss": 0.1316,
+      "step": 2387
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.695014662756599e-06,
+      "loss": 0.1302,
+      "step": 2388
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.694881364969342e-06,
+      "loss": 0.1772,
+      "step": 2389
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.694748067182086e-06,
+      "loss": 0.1004,
+      "step": 2390
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.69461476939483e-06,
+      "loss": 0.1022,
+      "step": 2391
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.694481471607572e-06,
+      "loss": 0.12,
+      "step": 2392
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.694348173820315e-06,
+      "loss": 0.1222,
+      "step": 2393
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.69421487603306e-06,
+      "loss": 0.1092,
+      "step": 2394
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.694081578245802e-06,
+      "loss": 0.117,
+      "step": 2395
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.693948280458545e-06,
+      "loss": 0.0963,
+      "step": 2396
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.693814982671288e-06,
+      "loss": 0.1107,
+      "step": 2397
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.693681684884032e-06,
+      "loss": 0.1583,
+      "step": 2398
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.693548387096775e-06,
+      "loss": 0.1098,
+      "step": 2399
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.693415089309518e-06,
+      "loss": 0.1299,
+      "step": 2400
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.69328179152226e-06,
+      "loss": 0.1174,
+      "step": 2401
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.693148493735005e-06,
+      "loss": 0.1274,
+      "step": 2402
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.693015195947748e-06,
+      "loss": 0.1326,
+      "step": 2403
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.692881898160491e-06,
+      "loss": 0.1111,
+      "step": 2404
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.692748600373234e-06,
+      "loss": 0.1211,
+      "step": 2405
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.692615302585978e-06,
+      "loss": 0.1154,
+      "step": 2406
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.692482004798721e-06,
+      "loss": 0.0936,
+      "step": 2407
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.692348707011464e-06,
+      "loss": 0.0862,
+      "step": 2408
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.692215409224207e-06,
+      "loss": 0.1432,
+      "step": 2409
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.692082111436951e-06,
+      "loss": 0.0839,
+      "step": 2410
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.691948813649694e-06,
+      "loss": 0.1292,
+      "step": 2411
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.691815515862437e-06,
+      "loss": 0.1309,
+      "step": 2412
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.691682218075181e-06,
+      "loss": 0.1396,
+      "step": 2413
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.691548920287924e-06,
+      "loss": 0.122,
+      "step": 2414
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.691415622500667e-06,
+      "loss": 0.1455,
+      "step": 2415
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.69128232471341e-06,
+      "loss": 0.1105,
+      "step": 2416
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.691149026926154e-06,
+      "loss": 0.1099,
+      "step": 2417
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.691015729138897e-06,
+      "loss": 0.1278,
+      "step": 2418
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.69088243135164e-06,
+      "loss": 0.1065,
+      "step": 2419
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.690749133564383e-06,
+      "loss": 0.1188,
+      "step": 2420
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.690615835777127e-06,
+      "loss": 0.1328,
+      "step": 2421
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.69048253798987e-06,
+      "loss": 0.1242,
+      "step": 2422
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.690349240202613e-06,
+      "loss": 0.1563,
+      "step": 2423
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.690215942415356e-06,
+      "loss": 0.1031,
+      "step": 2424
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.6900826446281e-06,
+      "loss": 0.1438,
+      "step": 2425
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.689949346840843e-06,
+      "loss": 0.1011,
+      "step": 2426
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.689816049053586e-06,
+      "loss": 0.1375,
+      "step": 2427
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.689682751266329e-06,
+      "loss": 0.1088,
+      "step": 2428
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.689549453479073e-06,
+      "loss": 0.1324,
+      "step": 2429
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.689416155691816e-06,
+      "loss": 0.1605,
+      "step": 2430
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.689282857904559e-06,
+      "loss": 0.1576,
+      "step": 2431
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.689149560117302e-06,
+      "loss": 0.146,
+      "step": 2432
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.689016262330046e-06,
+      "loss": 0.1257,
+      "step": 2433
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.688882964542789e-06,
+      "loss": 0.1344,
+      "step": 2434
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.688749666755532e-06,
+      "loss": 0.1017,
+      "step": 2435
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.688616368968276e-06,
+      "loss": 0.1412,
+      "step": 2436
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.68848307118102e-06,
+      "loss": 0.1103,
+      "step": 2437
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.688349773393762e-06,
+      "loss": 0.147,
+      "step": 2438
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.688216475606505e-06,
+      "loss": 0.128,
+      "step": 2439
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.68808317781925e-06,
+      "loss": 0.1409,
+      "step": 2440
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.687949880031992e-06,
+      "loss": 0.1265,
+      "step": 2441
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.687816582244735e-06,
+      "loss": 0.1199,
+      "step": 2442
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.68768328445748e-06,
+      "loss": 0.141,
+      "step": 2443
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.687549986670222e-06,
+      "loss": 0.1505,
+      "step": 2444
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.687416688882965e-06,
+      "loss": 0.1231,
+      "step": 2445
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.687283391095708e-06,
+      "loss": 0.1366,
+      "step": 2446
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.687150093308452e-06,
+      "loss": 0.1227,
+      "step": 2447
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.687016795521195e-06,
+      "loss": 0.1264,
+      "step": 2448
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.686883497733938e-06,
+      "loss": 0.0953,
+      "step": 2449
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.686750199946683e-06,
+      "loss": 0.1123,
+      "step": 2450
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.686616902159425e-06,
+      "loss": 0.124,
+      "step": 2451
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.686483604372168e-06,
+      "loss": 0.1039,
+      "step": 2452
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.686350306584911e-06,
+      "loss": 0.0966,
+      "step": 2453
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.686217008797656e-06,
+      "loss": 0.0945,
+      "step": 2454
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.686083711010398e-06,
+      "loss": 0.1286,
+      "step": 2455
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.685950413223141e-06,
+      "loss": 0.1031,
+      "step": 2456
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.685817115435886e-06,
+      "loss": 0.1481,
+      "step": 2457
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.685683817648629e-06,
+      "loss": 0.131,
+      "step": 2458
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.685550519861371e-06,
+      "loss": 0.0919,
+      "step": 2459
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.685417222074114e-06,
+      "loss": 0.1016,
+      "step": 2460
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.685283924286859e-06,
+      "loss": 0.1264,
+      "step": 2461
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.685150626499602e-06,
+      "loss": 0.1731,
+      "step": 2462
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.685017328712344e-06,
+      "loss": 0.149,
+      "step": 2463
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.684884030925087e-06,
+      "loss": 0.1187,
+      "step": 2464
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.684750733137832e-06,
+      "loss": 0.0924,
+      "step": 2465
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.684617435350574e-06,
+      "loss": 0.1264,
+      "step": 2466
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.684484137563317e-06,
+      "loss": 0.1148,
+      "step": 2467
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.684350839776062e-06,
+      "loss": 0.1698,
+      "step": 2468
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.684217541988805e-06,
+      "loss": 0.1387,
+      "step": 2469
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.684084244201547e-06,
+      "loss": 0.1262,
+      "step": 2470
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.68395094641429e-06,
+      "loss": 0.0939,
+      "step": 2471
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.683817648627035e-06,
+      "loss": 0.089,
+      "step": 2472
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.683684350839778e-06,
+      "loss": 0.1508,
+      "step": 2473
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.68355105305252e-06,
+      "loss": 0.1116,
+      "step": 2474
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.683417755265263e-06,
+      "loss": 0.1167,
+      "step": 2475
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.683284457478008e-06,
+      "loss": 0.1376,
+      "step": 2476
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.68315115969075e-06,
+      "loss": 0.1304,
+      "step": 2477
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.683017861903493e-06,
+      "loss": 0.1285,
+      "step": 2478
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.682884564116236e-06,
+      "loss": 0.1072,
+      "step": 2479
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.68275126632898e-06,
+      "loss": 0.1628,
+      "step": 2480
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.682617968541724e-06,
+      "loss": 0.1125,
+      "step": 2481
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.682484670754466e-06,
+      "loss": 0.1536,
+      "step": 2482
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.68235137296721e-06,
+      "loss": 0.1491,
+      "step": 2483
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.682218075179954e-06,
+      "loss": 0.1227,
+      "step": 2484
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.682084777392697e-06,
+      "loss": 0.145,
+      "step": 2485
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.68195147960544e-06,
+      "loss": 0.182,
+      "step": 2486
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.681818181818182e-06,
+      "loss": 0.1251,
+      "step": 2487
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.681684884030927e-06,
+      "loss": 0.1469,
+      "step": 2488
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.68155158624367e-06,
+      "loss": 0.1424,
+      "step": 2489
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.681418288456412e-06,
+      "loss": 0.1137,
+      "step": 2490
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.681284990669155e-06,
+      "loss": 0.1239,
+      "step": 2491
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.6811516928819e-06,
+      "loss": 0.1496,
+      "step": 2492
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.681018395094642e-06,
+      "loss": 0.1523,
+      "step": 2493
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.680885097307385e-06,
+      "loss": 0.1268,
+      "step": 2494
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.680751799520128e-06,
+      "loss": 0.1243,
+      "step": 2495
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.680618501732873e-06,
+      "loss": 0.0969,
+      "step": 2496
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.680485203945615e-06,
+      "loss": 0.1159,
+      "step": 2497
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.680351906158358e-06,
+      "loss": 0.1494,
+      "step": 2498
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.680218608371101e-06,
+      "loss": 0.1,
+      "step": 2499
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.680085310583846e-06,
+      "loss": 0.118,
+      "step": 2500
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.679977607762642e-06,
+      "loss": 0.1028,
+      "step": 2501
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.679844320635513e-06,
+      "loss": 0.1338,
+      "step": 2502
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.679711033508384e-06,
+      "loss": 0.1349,
+      "step": 2503
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.679577746381255e-06,
+      "loss": 0.0982,
+      "step": 2504
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.679444459254126e-06,
+      "loss": 0.1297,
+      "step": 2505
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.679311172126997e-06,
+      "loss": 0.1502,
+      "step": 2506
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.679177884999867e-06,
+      "loss": 0.1259,
+      "step": 2507
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.67904459787274e-06,
+      "loss": 0.1286,
+      "step": 2508
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.678911310745609e-06,
+      "loss": 0.1267,
+      "step": 2509
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.67877802361848e-06,
+      "loss": 0.1256,
+      "step": 2510
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.67864473649135e-06,
+      "loss": 0.0817,
+      "step": 2511
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.67851144936422e-06,
+      "loss": 0.1014,
+      "step": 2512
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.678378162237093e-06,
+      "loss": 0.1521,
+      "step": 2513
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.678244875109962e-06,
+      "loss": 0.1212,
+      "step": 2514
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.678111587982833e-06,
+      "loss": 0.1235,
+      "step": 2515
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.677978300855704e-06,
+      "loss": 0.1469,
+      "step": 2516
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.677845013728575e-06,
+      "loss": 0.1264,
+      "step": 2517
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.677711726601446e-06,
+      "loss": 0.1223,
+      "step": 2518
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.677578439474317e-06,
+      "loss": 0.1248,
+      "step": 2519
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.677445152347186e-06,
+      "loss": 0.0948,
+      "step": 2520
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.677311865220059e-06,
+      "loss": 0.1079,
+      "step": 2521
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.677178578092928e-06,
+      "loss": 0.1051,
+      "step": 2522
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.6770452909658e-06,
+      "loss": 0.1234,
+      "step": 2523
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.67691200383867e-06,
+      "loss": 0.1031,
+      "step": 2524
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.676778716711541e-06,
+      "loss": 0.1776,
+      "step": 2525
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.676645429584412e-06,
+      "loss": 0.135,
+      "step": 2526
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.676512142457283e-06,
+      "loss": 0.1418,
+      "step": 2527
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.676378855330153e-06,
+      "loss": 0.1415,
+      "step": 2528
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.676245568203024e-06,
+      "loss": 0.0914,
+      "step": 2529
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.676112281075895e-06,
+      "loss": 0.1372,
+      "step": 2530
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.675978993948766e-06,
+      "loss": 0.09,
+      "step": 2531
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.675845706821637e-06,
+      "loss": 0.1206,
+      "step": 2532
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.675712419694506e-06,
+      "loss": 0.1214,
+      "step": 2533
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.675579132567377e-06,
+      "loss": 0.1523,
+      "step": 2534
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.675445845440248e-06,
+      "loss": 0.0985,
+      "step": 2535
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.675312558313119e-06,
+      "loss": 0.1104,
+      "step": 2536
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.67517927118599e-06,
+      "loss": 0.1133,
+      "step": 2537
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.67504598405886e-06,
+      "loss": 0.1225,
+      "step": 2538
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.67491269693173e-06,
+      "loss": 0.0875,
+      "step": 2539
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.674779409804603e-06,
+      "loss": 0.1395,
+      "step": 2540
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.674646122677472e-06,
+      "loss": 0.1228,
+      "step": 2541
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.674512835550343e-06,
+      "loss": 0.1279,
+      "step": 2542
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.674379548423214e-06,
+      "loss": 0.1314,
+      "step": 2543
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.674246261296085e-06,
+      "loss": 0.1411,
+      "step": 2544
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.674112974168956e-06,
+      "loss": 0.1043,
+      "step": 2545
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.673979687041827e-06,
+      "loss": 0.1121,
+      "step": 2546
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.673846399914696e-06,
+      "loss": 0.1204,
+      "step": 2547
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.673713112787569e-06,
+      "loss": 0.1217,
+      "step": 2548
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.673579825660438e-06,
+      "loss": 0.1181,
+      "step": 2549
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.67344653853331e-06,
+      "loss": 0.1112,
+      "step": 2550
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.67331325140618e-06,
+      "loss": 0.0999,
+      "step": 2551
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.67317996427905e-06,
+      "loss": 0.1303,
+      "step": 2552
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.673046677151922e-06,
+      "loss": 0.1425,
+      "step": 2553
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.672913390024792e-06,
+      "loss": 0.1425,
+      "step": 2554
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.672780102897663e-06,
+      "loss": 0.0989,
+      "step": 2555
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.672646815770534e-06,
+      "loss": 0.1312,
+      "step": 2556
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.672513528643405e-06,
+      "loss": 0.1256,
+      "step": 2557
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.672380241516276e-06,
+      "loss": 0.0756,
+      "step": 2558
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.672246954389147e-06,
+      "loss": 0.0883,
+      "step": 2559
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.672113667262016e-06,
+      "loss": 0.1152,
+      "step": 2560
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.671980380134887e-06,
+      "loss": 0.1745,
+      "step": 2561
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.671847093007758e-06,
+      "loss": 0.069,
+      "step": 2562
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.671713805880629e-06,
+      "loss": 0.1237,
+      "step": 2563
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.6715805187535e-06,
+      "loss": 0.1041,
+      "step": 2564
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.671447231626371e-06,
+      "loss": 0.1237,
+      "step": 2565
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.67131394449924e-06,
+      "loss": 0.121,
+      "step": 2566
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.671180657372113e-06,
+      "loss": 0.107,
+      "step": 2567
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.671047370244982e-06,
+      "loss": 0.1311,
+      "step": 2568
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.670914083117853e-06,
+      "loss": 0.1127,
+      "step": 2569
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.670780795990724e-06,
+      "loss": 0.1052,
+      "step": 2570
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.670647508863593e-06,
+      "loss": 0.1236,
+      "step": 2571
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.670514221736466e-06,
+      "loss": 0.1122,
+      "step": 2572
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.670380934609335e-06,
+      "loss": 0.1811,
+      "step": 2573
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.670247647482206e-06,
+      "loss": 0.1365,
+      "step": 2574
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.670114360355077e-06,
+      "loss": 0.1058,
+      "step": 2575
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.669981073227948e-06,
+      "loss": 0.1149,
+      "step": 2576
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.66984778610082e-06,
+      "loss": 0.1152,
+      "step": 2577
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.66971449897369e-06,
+      "loss": 0.1317,
+      "step": 2578
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.66958121184656e-06,
+      "loss": 0.1804,
+      "step": 2579
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.669447924719432e-06,
+      "loss": 0.1458,
+      "step": 2580
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.669314637592302e-06,
+      "loss": 0.0738,
+      "step": 2581
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.669181350465173e-06,
+      "loss": 0.1077,
+      "step": 2582
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.669048063338044e-06,
+      "loss": 0.1308,
+      "step": 2583
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.668914776210915e-06,
+      "loss": 0.1011,
+      "step": 2584
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.668781489083786e-06,
+      "loss": 0.0956,
+      "step": 2585
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.668648201956657e-06,
+      "loss": 0.1632,
+      "step": 2586
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.668514914829526e-06,
+      "loss": 0.1233,
+      "step": 2587
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.668381627702397e-06,
+      "loss": 0.1322,
+      "step": 2588
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.668248340575268e-06,
+      "loss": 0.1068,
+      "step": 2589
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.668115053448139e-06,
+      "loss": 0.108,
+      "step": 2590
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.66798176632101e-06,
+      "loss": 0.1404,
+      "step": 2591
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.66784847919388e-06,
+      "loss": 0.1212,
+      "step": 2592
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.667715192066752e-06,
+      "loss": 0.1317,
+      "step": 2593
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.667581904939621e-06,
+      "loss": 0.1455,
+      "step": 2594
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.667448617812492e-06,
+      "loss": 0.1543,
+      "step": 2595
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.667315330685363e-06,
+      "loss": 0.0841,
+      "step": 2596
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.667182043558234e-06,
+      "loss": 0.148,
+      "step": 2597
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.667048756431105e-06,
+      "loss": 0.1148,
+      "step": 2598
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.666915469303976e-06,
+      "loss": 0.0847,
+      "step": 2599
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.666782182176846e-06,
+      "loss": 0.1309,
+      "step": 2600
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.666648895049717e-06,
+      "loss": 0.125,
+      "step": 2601
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.666515607922588e-06,
+      "loss": 0.1285,
+      "step": 2602
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.666382320795459e-06,
+      "loss": 0.116,
+      "step": 2603
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.66624903366833e-06,
+      "loss": 0.1123,
+      "step": 2604
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.6661157465412e-06,
+      "loss": 0.1311,
+      "step": 2605
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.66598245941407e-06,
+      "loss": 0.1348,
+      "step": 2606
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.665849172286942e-06,
+      "loss": 0.1168,
+      "step": 2607
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.665715885159812e-06,
+      "loss": 0.1247,
+      "step": 2608
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.665582598032683e-06,
+      "loss": 0.0976,
+      "step": 2609
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.665449310905554e-06,
+      "loss": 0.1209,
+      "step": 2610
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.665316023778423e-06,
+      "loss": 0.1655,
+      "step": 2611
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.665182736651296e-06,
+      "loss": 0.1112,
+      "step": 2612
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.665049449524165e-06,
+      "loss": 0.1237,
+      "step": 2613
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.664916162397036e-06,
+      "loss": 0.1378,
+      "step": 2614
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.664782875269907e-06,
+      "loss": 0.1597,
+      "step": 2615
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.664649588142778e-06,
+      "loss": 0.1195,
+      "step": 2616
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.664516301015649e-06,
+      "loss": 0.0984,
+      "step": 2617
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.66438301388852e-06,
+      "loss": 0.1194,
+      "step": 2618
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.66424972676139e-06,
+      "loss": 0.1,
+      "step": 2619
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.664116439634262e-06,
+      "loss": 0.1721,
+      "step": 2620
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.663983152507131e-06,
+      "loss": 0.1138,
+      "step": 2621
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.663849865380002e-06,
+      "loss": 0.1061,
+      "step": 2622
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.663716578252873e-06,
+      "loss": 0.1218,
+      "step": 2623
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.663583291125744e-06,
+      "loss": 0.107,
+      "step": 2624
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.663450003998615e-06,
+      "loss": 0.0953,
+      "step": 2625
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.663316716871486e-06,
+      "loss": 0.1272,
+      "step": 2626
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.663183429744356e-06,
+      "loss": 0.14,
+      "step": 2627
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.663050142617227e-06,
+      "loss": 0.1603,
+      "step": 2628
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.662916855490098e-06,
+      "loss": 0.1583,
+      "step": 2629
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.662783568362969e-06,
+      "loss": 0.1194,
+      "step": 2630
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.66265028123584e-06,
+      "loss": 0.1352,
+      "step": 2631
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.662516994108709e-06,
+      "loss": 0.1048,
+      "step": 2632
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.66238370698158e-06,
+      "loss": 0.1027,
+      "step": 2633
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.662250419854451e-06,
+      "loss": 0.0896,
+      "step": 2634
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.662117132727322e-06,
+      "loss": 0.1172,
+      "step": 2635
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.661983845600193e-06,
+      "loss": 0.0974,
+      "step": 2636
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.661850558473064e-06,
+      "loss": 0.1064,
+      "step": 2637
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.661717271345933e-06,
+      "loss": 0.1019,
+      "step": 2638
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.661583984218806e-06,
+      "loss": 0.1385,
+      "step": 2639
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.661450697091675e-06,
+      "loss": 0.1075,
+      "step": 2640
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.661317409964546e-06,
+      "loss": 0.1226,
+      "step": 2641
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.661184122837417e-06,
+      "loss": 0.1084,
+      "step": 2642
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.661050835710288e-06,
+      "loss": 0.1254,
+      "step": 2643
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.660917548583159e-06,
+      "loss": 0.1305,
+      "step": 2644
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.66078426145603e-06,
+      "loss": 0.1247,
+      "step": 2645
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.6606509743289e-06,
+      "loss": 0.1685,
+      "step": 2646
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.66051768720177e-06,
+      "loss": 0.136,
+      "step": 2647
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.660384400074641e-06,
+      "loss": 0.0965,
+      "step": 2648
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.660251112947512e-06,
+      "loss": 0.1596,
+      "step": 2649
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.660117825820383e-06,
+      "loss": 0.1403,
+      "step": 2650
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.659984538693253e-06,
+      "loss": 0.1205,
+      "step": 2651
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.659851251566125e-06,
+      "loss": 0.1384,
+      "step": 2652
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.659717964438995e-06,
+      "loss": 0.1107,
+      "step": 2653
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.659584677311866e-06,
+      "loss": 0.1366,
+      "step": 2654
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.659451390184737e-06,
+      "loss": 0.1027,
+      "step": 2655
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.659318103057608e-06,
+      "loss": 0.1318,
+      "step": 2656
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.659184815930479e-06,
+      "loss": 0.143,
+      "step": 2657
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.65905152880335e-06,
+      "loss": 0.1873,
+      "step": 2658
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.658918241676219e-06,
+      "loss": 0.0923,
+      "step": 2659
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.658784954549092e-06,
+      "loss": 0.1334,
+      "step": 2660
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.658651667421961e-06,
+      "loss": 0.1038,
+      "step": 2661
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.658518380294832e-06,
+      "loss": 0.1282,
+      "step": 2662
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.658385093167703e-06,
+      "loss": 0.119,
+      "step": 2663
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.658251806040574e-06,
+      "loss": 0.1237,
+      "step": 2664
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.658118518913445e-06,
+      "loss": 0.0961,
+      "step": 2665
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.657985231786316e-06,
+      "loss": 0.1044,
+      "step": 2666
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.657851944659185e-06,
+      "loss": 0.107,
+      "step": 2667
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.657718657532056e-06,
+      "loss": 0.1449,
+      "step": 2668
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.657585370404927e-06,
+      "loss": 0.1017,
+      "step": 2669
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.657452083277797e-06,
+      "loss": 0.1077,
+      "step": 2670
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.65731879615067e-06,
+      "loss": 0.1139,
+      "step": 2671
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.657185509023539e-06,
+      "loss": 0.1489,
+      "step": 2672
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.65705222189641e-06,
+      "loss": 0.1618,
+      "step": 2673
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.65691893476928e-06,
+      "loss": 0.1702,
+      "step": 2674
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.656785647642152e-06,
+      "loss": 0.1606,
+      "step": 2675
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.656652360515022e-06,
+      "loss": 0.1439,
+      "step": 2676
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.656519073387893e-06,
+      "loss": 0.093,
+      "step": 2677
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.656385786260763e-06,
+      "loss": 0.1269,
+      "step": 2678
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.656252499133635e-06,
+      "loss": 0.1302,
+      "step": 2679
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.656119212006505e-06,
+      "loss": 0.1153,
+      "step": 2680
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.655985924879376e-06,
+      "loss": 0.0822,
+      "step": 2681
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.655852637752247e-06,
+      "loss": 0.1256,
+      "step": 2682
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.655719350625118e-06,
+      "loss": 0.1444,
+      "step": 2683
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.655586063497989e-06,
+      "loss": 0.1342,
+      "step": 2684
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.65545277637086e-06,
+      "loss": 0.1698,
+      "step": 2685
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.655319489243729e-06,
+      "loss": 0.1206,
+      "step": 2686
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.6551862021166e-06,
+      "loss": 0.0842,
+      "step": 2687
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.655052914989471e-06,
+      "loss": 0.1669,
+      "step": 2688
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.654919627862342e-06,
+      "loss": 0.1067,
+      "step": 2689
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.654786340735213e-06,
+      "loss": 0.1066,
+      "step": 2690
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.654653053608082e-06,
+      "loss": 0.1394,
+      "step": 2691
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.654519766480955e-06,
+      "loss": 0.1152,
+      "step": 2692
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.654386479353824e-06,
+      "loss": 0.1274,
+      "step": 2693
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.654253192226695e-06,
+      "loss": 0.1385,
+      "step": 2694
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.654119905099566e-06,
+      "loss": 0.1088,
+      "step": 2695
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.653986617972437e-06,
+      "loss": 0.1418,
+      "step": 2696
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.653853330845308e-06,
+      "loss": 0.1588,
+      "step": 2697
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.65372004371818e-06,
+      "loss": 0.1274,
+      "step": 2698
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.653586756591049e-06,
+      "loss": 0.1632,
+      "step": 2699
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.65345346946392e-06,
+      "loss": 0.0996,
+      "step": 2700
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.65332018233679e-06,
+      "loss": 0.0899,
+      "step": 2701
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.653186895209662e-06,
+      "loss": 0.1086,
+      "step": 2702
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.653053608082533e-06,
+      "loss": 0.1366,
+      "step": 2703
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.652920320955404e-06,
+      "loss": 0.1056,
+      "step": 2704
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.652787033828273e-06,
+      "loss": 0.1278,
+      "step": 2705
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.652653746701144e-06,
+      "loss": 0.2091,
+      "step": 2706
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.652520459574015e-06,
+      "loss": 0.0965,
+      "step": 2707
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.652387172446886e-06,
+      "loss": 0.1623,
+      "step": 2708
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.652253885319757e-06,
+      "loss": 0.133,
+      "step": 2709
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.652120598192626e-06,
+      "loss": 0.1384,
+      "step": 2710
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.651987311065499e-06,
+      "loss": 0.1277,
+      "step": 2711
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.651854023938368e-06,
+      "loss": 0.1699,
+      "step": 2712
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.651720736811239e-06,
+      "loss": 0.1501,
+      "step": 2713
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.65158744968411e-06,
+      "loss": 0.174,
+      "step": 2714
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.651454162556981e-06,
+      "loss": 0.1115,
+      "step": 2715
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.651320875429852e-06,
+      "loss": 0.1249,
+      "step": 2716
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.651187588302723e-06,
+      "loss": 0.1202,
+      "step": 2717
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.651054301175592e-06,
+      "loss": 0.113,
+      "step": 2718
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.650921014048465e-06,
+      "loss": 0.076,
+      "step": 2719
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.650787726921334e-06,
+      "loss": 0.099,
+      "step": 2720
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.650654439794205e-06,
+      "loss": 0.1554,
+      "step": 2721
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.650521152667076e-06,
+      "loss": 0.1487,
+      "step": 2722
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.650387865539947e-06,
+      "loss": 0.1224,
+      "step": 2723
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.650254578412818e-06,
+      "loss": 0.0899,
+      "step": 2724
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.650121291285688e-06,
+      "loss": 0.1292,
+      "step": 2725
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.649988004158559e-06,
+      "loss": 0.1045,
+      "step": 2726
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.64985471703143e-06,
+      "loss": 0.1002,
+      "step": 2727
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.6497214299043e-06,
+      "loss": 0.1473,
+      "step": 2728
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.649588142777172e-06,
+      "loss": 0.1435,
+      "step": 2729
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.649454855650043e-06,
+      "loss": 0.1099,
+      "step": 2730
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.649321568522912e-06,
+      "loss": 0.0937,
+      "step": 2731
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.649188281395783e-06,
+      "loss": 0.1254,
+      "step": 2732
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.649054994268654e-06,
+      "loss": 0.1777,
+      "step": 2733
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.648921707141525e-06,
+      "loss": 0.1296,
+      "step": 2734
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.648788420014396e-06,
+      "loss": 0.1016,
+      "step": 2735
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.648655132887267e-06,
+      "loss": 0.1285,
+      "step": 2736
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.648521845760136e-06,
+      "loss": 0.1334,
+      "step": 2737
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.648388558633009e-06,
+      "loss": 0.1241,
+      "step": 2738
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.648255271505878e-06,
+      "loss": 0.1187,
+      "step": 2739
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.64812198437875e-06,
+      "loss": 0.1305,
+      "step": 2740
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.64798869725162e-06,
+      "loss": 0.1022,
+      "step": 2741
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.647855410124491e-06,
+      "loss": 0.1209,
+      "step": 2742
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.647722122997362e-06,
+      "loss": 0.1688,
+      "step": 2743
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.647588835870233e-06,
+      "loss": 0.132,
+      "step": 2744
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.647455548743102e-06,
+      "loss": 0.1223,
+      "step": 2745
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.647322261615973e-06,
+      "loss": 0.1168,
+      "step": 2746
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.647188974488844e-06,
+      "loss": 0.1173,
+      "step": 2747
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.647055687361715e-06,
+      "loss": 0.1235,
+      "step": 2748
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.646922400234586e-06,
+      "loss": 0.1084,
+      "step": 2749
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.646789113107456e-06,
+      "loss": 0.0767,
+      "step": 2750
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.646655825980328e-06,
+      "loss": 0.1402,
+      "step": 2751
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.646522538853198e-06,
+      "loss": 0.1238,
+      "step": 2752
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.646389251726069e-06,
+      "loss": 0.1496,
+      "step": 2753
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.64625596459894e-06,
+      "loss": 0.1232,
+      "step": 2754
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.64612267747181e-06,
+      "loss": 0.1352,
+      "step": 2755
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.645989390344682e-06,
+      "loss": 0.0772,
+      "step": 2756
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.645856103217553e-06,
+      "loss": 0.1321,
+      "step": 2757
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.645722816090422e-06,
+      "loss": 0.1264,
+      "step": 2758
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.645589528963295e-06,
+      "loss": 0.0939,
+      "step": 2759
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.645456241836164e-06,
+      "loss": 0.1489,
+      "step": 2760
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.645322954709035e-06,
+      "loss": 0.0825,
+      "step": 2761
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.645189667581906e-06,
+      "loss": 0.0886,
+      "step": 2762
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.645056380454777e-06,
+      "loss": 0.1005,
+      "step": 2763
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.644923093327648e-06,
+      "loss": 0.1449,
+      "step": 2764
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.644789806200517e-06,
+      "loss": 0.173,
+      "step": 2765
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.644656519073388e-06,
+      "loss": 0.1454,
+      "step": 2766
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.64452323194626e-06,
+      "loss": 0.1016,
+      "step": 2767
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.64438994481913e-06,
+      "loss": 0.1079,
+      "step": 2768
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.644256657692001e-06,
+      "loss": 0.1042,
+      "step": 2769
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.644123370564872e-06,
+      "loss": 0.1171,
+      "step": 2770
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.643990083437742e-06,
+      "loss": 0.1254,
+      "step": 2771
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.643856796310613e-06,
+      "loss": 0.091,
+      "step": 2772
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.643723509183484e-06,
+      "loss": 0.129,
+      "step": 2773
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.643590222056355e-06,
+      "loss": 0.1071,
+      "step": 2774
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.643456934929226e-06,
+      "loss": 0.1131,
+      "step": 2775
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.643323647802097e-06,
+      "loss": 0.1186,
+      "step": 2776
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.643190360674966e-06,
+      "loss": 0.1202,
+      "step": 2777
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.643057073547839e-06,
+      "loss": 0.1051,
+      "step": 2778
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.642923786420708e-06,
+      "loss": 0.1467,
+      "step": 2779
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.642790499293579e-06,
+      "loss": 0.1158,
+      "step": 2780
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.64265721216645e-06,
+      "loss": 0.1027,
+      "step": 2781
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.64252392503932e-06,
+      "loss": 0.1438,
+      "step": 2782
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.642390637912192e-06,
+      "loss": 0.1173,
+      "step": 2783
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.642257350785061e-06,
+      "loss": 0.1324,
+      "step": 2784
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.642124063657932e-06,
+      "loss": 0.0922,
+      "step": 2785
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.641990776530803e-06,
+      "loss": 0.1327,
+      "step": 2786
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.641857489403674e-06,
+      "loss": 0.1178,
+      "step": 2787
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.641724202276545e-06,
+      "loss": 0.0882,
+      "step": 2788
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.641590915149416e-06,
+      "loss": 0.1305,
+      "step": 2789
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.641457628022285e-06,
+      "loss": 0.1492,
+      "step": 2790
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.641324340895158e-06,
+      "loss": 0.1172,
+      "step": 2791
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.641191053768027e-06,
+      "loss": 0.0947,
+      "step": 2792
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.641057766640898e-06,
+      "loss": 0.1324,
+      "step": 2793
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.64092447951377e-06,
+      "loss": 0.123,
+      "step": 2794
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.64079119238664e-06,
+      "loss": 0.1284,
+      "step": 2795
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.640657905259511e-06,
+      "loss": 0.1038,
+      "step": 2796
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.640524618132382e-06,
+      "loss": 0.1363,
+      "step": 2797
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.640391331005252e-06,
+      "loss": 0.1149,
+      "step": 2798
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.640258043878123e-06,
+      "loss": 0.1627,
+      "step": 2799
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.640124756750994e-06,
+      "loss": 0.088,
+      "step": 2800
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.639991469623865e-06,
+      "loss": 0.158,
+      "step": 2801
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.639858182496736e-06,
+      "loss": 0.1672,
+      "step": 2802
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.639724895369607e-06,
+      "loss": 0.1288,
+      "step": 2803
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.639591608242476e-06,
+      "loss": 0.1495,
+      "step": 2804
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.639458321115347e-06,
+      "loss": 0.175,
+      "step": 2805
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.639325033988218e-06,
+      "loss": 0.1299,
+      "step": 2806
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.639191746861089e-06,
+      "loss": 0.0827,
+      "step": 2807
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.63905845973396e-06,
+      "loss": 0.1118,
+      "step": 2808
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.63892517260683e-06,
+      "loss": 0.1092,
+      "step": 2809
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.638791885479702e-06,
+      "loss": 0.1561,
+      "step": 2810
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.638658598352571e-06,
+      "loss": 0.1331,
+      "step": 2811
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.638525311225442e-06,
+      "loss": 0.1267,
+      "step": 2812
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.638392024098313e-06,
+      "loss": 0.0908,
+      "step": 2813
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.638258736971184e-06,
+      "loss": 0.1394,
+      "step": 2814
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.638125449844055e-06,
+      "loss": 0.1391,
+      "step": 2815
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.637992162716926e-06,
+      "loss": 0.1696,
+      "step": 2816
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.637858875589795e-06,
+      "loss": 0.1505,
+      "step": 2817
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.637725588462668e-06,
+      "loss": 0.1061,
+      "step": 2818
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.637592301335537e-06,
+      "loss": 0.138,
+      "step": 2819
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.637459014208408e-06,
+      "loss": 0.1189,
+      "step": 2820
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.63732572708128e-06,
+      "loss": 0.0893,
+      "step": 2821
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.63719243995415e-06,
+      "loss": 0.1202,
+      "step": 2822
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.637059152827021e-06,
+      "loss": 0.1077,
+      "step": 2823
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.63692586569989e-06,
+      "loss": 0.1494,
+      "step": 2824
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.636792578572762e-06,
+      "loss": 0.1354,
+      "step": 2825
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.636659291445633e-06,
+      "loss": 0.1206,
+      "step": 2826
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.636526004318504e-06,
+      "loss": 0.093,
+      "step": 2827
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.636392717191375e-06,
+      "loss": 0.1022,
+      "step": 2828
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.636259430064246e-06,
+      "loss": 0.1132,
+      "step": 2829
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.636126142937115e-06,
+      "loss": 0.1367,
+      "step": 2830
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.635992855809988e-06,
+      "loss": 0.1197,
+      "step": 2831
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.635859568682857e-06,
+      "loss": 0.1003,
+      "step": 2832
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.635726281555728e-06,
+      "loss": 0.1183,
+      "step": 2833
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.635592994428599e-06,
+      "loss": 0.1273,
+      "step": 2834
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.63545970730147e-06,
+      "loss": 0.1134,
+      "step": 2835
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.635326420174341e-06,
+      "loss": 0.1863,
+      "step": 2836
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.635193133047212e-06,
+      "loss": 0.128,
+      "step": 2837
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.635059845920081e-06,
+      "loss": 0.136,
+      "step": 2838
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.634926558792952e-06,
+      "loss": 0.1043,
+      "step": 2839
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.634793271665823e-06,
+      "loss": 0.1425,
+      "step": 2840
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.634659984538694e-06,
+      "loss": 0.1012,
+      "step": 2841
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.634526697411565e-06,
+      "loss": 0.1192,
+      "step": 2842
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.634393410284435e-06,
+      "loss": 0.1174,
+      "step": 2843
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.634260123157306e-06,
+      "loss": 0.1288,
+      "step": 2844
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.634126836030177e-06,
+      "loss": 0.1116,
+      "step": 2845
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.633993548903048e-06,
+      "loss": 0.1165,
+      "step": 2846
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.633860261775919e-06,
+      "loss": 0.1049,
+      "step": 2847
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.63372697464879e-06,
+      "loss": 0.1573,
+      "step": 2848
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.633593687521659e-06,
+      "loss": 0.1338,
+      "step": 2849
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.633460400394532e-06,
+      "loss": 0.1057,
+      "step": 2850
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.6333271132674e-06,
+      "loss": 0.0895,
+      "step": 2851
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.633193826140272e-06,
+      "loss": 0.1103,
+      "step": 2852
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.633060539013143e-06,
+      "loss": 0.097,
+      "step": 2853
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.632927251886014e-06,
+      "loss": 0.1374,
+      "step": 2854
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.632793964758885e-06,
+      "loss": 0.1261,
+      "step": 2855
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.632660677631756e-06,
+      "loss": 0.11,
+      "step": 2856
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.632527390504625e-06,
+      "loss": 0.1253,
+      "step": 2857
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.632394103377498e-06,
+      "loss": 0.1421,
+      "step": 2858
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.632260816250367e-06,
+      "loss": 0.0851,
+      "step": 2859
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.632127529123238e-06,
+      "loss": 0.1087,
+      "step": 2860
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.631994241996109e-06,
+      "loss": 0.1146,
+      "step": 2861
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.63186095486898e-06,
+      "loss": 0.1025,
+      "step": 2862
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.631727667741851e-06,
+      "loss": 0.0913,
+      "step": 2863
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.63159438061472e-06,
+      "loss": 0.1189,
+      "step": 2864
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.631461093487591e-06,
+      "loss": 0.1468,
+      "step": 2865
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.631327806360462e-06,
+      "loss": 0.108,
+      "step": 2866
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.631194519233333e-06,
+      "loss": 0.0933,
+      "step": 2867
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.631061232106204e-06,
+      "loss": 0.1506,
+      "step": 2868
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.630927944979075e-06,
+      "loss": 0.1116,
+      "step": 2869
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.630794657851945e-06,
+      "loss": 0.119,
+      "step": 2870
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.630661370724816e-06,
+      "loss": 0.1353,
+      "step": 2871
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.630528083597687e-06,
+      "loss": 0.1064,
+      "step": 2872
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.630394796470558e-06,
+      "loss": 0.1204,
+      "step": 2873
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.630261509343429e-06,
+      "loss": 0.1704,
+      "step": 2874
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.6301282222163e-06,
+      "loss": 0.1378,
+      "step": 2875
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.629994935089169e-06,
+      "loss": 0.124,
+      "step": 2876
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.629861647962042e-06,
+      "loss": 0.1071,
+      "step": 2877
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.629728360834911e-06,
+      "loss": 0.1125,
+      "step": 2878
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.629595073707782e-06,
+      "loss": 0.1344,
+      "step": 2879
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.629461786580653e-06,
+      "loss": 0.1089,
+      "step": 2880
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.629328499453524e-06,
+      "loss": 0.128,
+      "step": 2881
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.629195212326395e-06,
+      "loss": 0.1271,
+      "step": 2882
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.629061925199264e-06,
+      "loss": 0.1379,
+      "step": 2883
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.628928638072135e-06,
+      "loss": 0.1077,
+      "step": 2884
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.628795350945006e-06,
+      "loss": 0.1254,
+      "step": 2885
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.628662063817877e-06,
+      "loss": 0.141,
+      "step": 2886
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.628528776690748e-06,
+      "loss": 0.1176,
+      "step": 2887
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.62839548956362e-06,
+      "loss": 0.1297,
+      "step": 2888
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.628262202436488e-06,
+      "loss": 0.1213,
+      "step": 2889
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.628128915309361e-06,
+      "loss": 0.1248,
+      "step": 2890
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.62799562818223e-06,
+      "loss": 0.1116,
+      "step": 2891
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.627862341055101e-06,
+      "loss": 0.1541,
+      "step": 2892
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.627729053927972e-06,
+      "loss": 0.1277,
+      "step": 2893
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.627595766800843e-06,
+      "loss": 0.1285,
+      "step": 2894
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.627462479673714e-06,
+      "loss": 0.1374,
+      "step": 2895
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.627329192546585e-06,
+      "loss": 0.1329,
+      "step": 2896
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.627195905419455e-06,
+      "loss": 0.1248,
+      "step": 2897
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.627062618292327e-06,
+      "loss": 0.0983,
+      "step": 2898
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.626929331165197e-06,
+      "loss": 0.14,
+      "step": 2899
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.626796044038068e-06,
+      "loss": 0.0901,
+      "step": 2900
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.626662756910939e-06,
+      "loss": 0.1171,
+      "step": 2901
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.626529469783808e-06,
+      "loss": 0.224,
+      "step": 2902
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.62639618265668e-06,
+      "loss": 0.1487,
+      "step": 2903
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.62626289552955e-06,
+      "loss": 0.1531,
+      "step": 2904
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.626129608402421e-06,
+      "loss": 0.1089,
+      "step": 2905
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.625996321275292e-06,
+      "loss": 0.1711,
+      "step": 2906
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.625863034148163e-06,
+      "loss": 0.1259,
+      "step": 2907
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.625729747021034e-06,
+      "loss": 0.0944,
+      "step": 2908
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.625596459893905e-06,
+      "loss": 0.087,
+      "step": 2909
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.625463172766774e-06,
+      "loss": 0.1201,
+      "step": 2910
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.625329885639645e-06,
+      "loss": 0.1111,
+      "step": 2911
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.625196598512516e-06,
+      "loss": 0.1118,
+      "step": 2912
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.625063311385387e-06,
+      "loss": 0.1692,
+      "step": 2913
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.624930024258258e-06,
+      "loss": 0.1174,
+      "step": 2914
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.62479673713113e-06,
+      "loss": 0.1211,
+      "step": 2915
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.624663450003999e-06,
+      "loss": 0.1655,
+      "step": 2916
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.624530162876871e-06,
+      "loss": 0.1288,
+      "step": 2917
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.62439687574974e-06,
+      "loss": 0.0905,
+      "step": 2918
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.624263588622612e-06,
+      "loss": 0.1103,
+      "step": 2919
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.624130301495483e-06,
+      "loss": 0.1142,
+      "step": 2920
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.623997014368352e-06,
+      "loss": 0.0854,
+      "step": 2921
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.623863727241225e-06,
+      "loss": 0.1133,
+      "step": 2922
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.623730440114094e-06,
+      "loss": 0.0765,
+      "step": 2923
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.623597152986965e-06,
+      "loss": 0.1108,
+      "step": 2924
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.623463865859836e-06,
+      "loss": 0.1331,
+      "step": 2925
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.623330578732707e-06,
+      "loss": 0.1208,
+      "step": 2926
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.623197291605578e-06,
+      "loss": 0.1389,
+      "step": 2927
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.623064004478449e-06,
+      "loss": 0.0961,
+      "step": 2928
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.622930717351318e-06,
+      "loss": 0.1349,
+      "step": 2929
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.62279743022419e-06,
+      "loss": 0.1153,
+      "step": 2930
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.62266414309706e-06,
+      "loss": 0.1155,
+      "step": 2931
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.622530855969931e-06,
+      "loss": 0.1419,
+      "step": 2932
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.622397568842802e-06,
+      "loss": 0.124,
+      "step": 2933
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.622264281715673e-06,
+      "loss": 0.1126,
+      "step": 2934
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.622130994588544e-06,
+      "loss": 0.1371,
+      "step": 2935
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.621997707461415e-06,
+      "loss": 0.1041,
+      "step": 2936
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.621864420334284e-06,
+      "loss": 0.1171,
+      "step": 2937
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.621731133207155e-06,
+      "loss": 0.1332,
+      "step": 2938
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.621597846080026e-06,
+      "loss": 0.154,
+      "step": 2939
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.621464558952897e-06,
+      "loss": 0.1084,
+      "step": 2940
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.621331271825768e-06,
+      "loss": 0.1305,
+      "step": 2941
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.621197984698638e-06,
+      "loss": 0.1557,
+      "step": 2942
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.621064697571509e-06,
+      "loss": 0.1039,
+      "step": 2943
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.62093141044438e-06,
+      "loss": 0.1648,
+      "step": 2944
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.62079812331725e-06,
+      "loss": 0.1242,
+      "step": 2945
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.620664836190122e-06,
+      "loss": 0.0929,
+      "step": 2946
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.620531549062993e-06,
+      "loss": 0.1499,
+      "step": 2947
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.620398261935862e-06,
+      "loss": 0.1124,
+      "step": 2948
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.620264974808735e-06,
+      "loss": 0.1062,
+      "step": 2949
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.620131687681604e-06,
+      "loss": 0.1652,
+      "step": 2950
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.619998400554475e-06,
+      "loss": 0.0926,
+      "step": 2951
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.619865113427346e-06,
+      "loss": 0.1388,
+      "step": 2952
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.619731826300217e-06,
+      "loss": 0.1356,
+      "step": 2953
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.619598539173088e-06,
+      "loss": 0.1352,
+      "step": 2954
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.619465252045959e-06,
+      "loss": 0.1019,
+      "step": 2955
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.619331964918828e-06,
+      "loss": 0.1471,
+      "step": 2956
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.619198677791701e-06,
+      "loss": 0.111,
+      "step": 2957
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.61906539066457e-06,
+      "loss": 0.092,
+      "step": 2958
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.618932103537441e-06,
+      "loss": 0.132,
+      "step": 2959
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.618798816410312e-06,
+      "loss": 0.0965,
+      "step": 2960
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.618665529283181e-06,
+      "loss": 0.1689,
+      "step": 2961
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.618532242156054e-06,
+      "loss": 0.1282,
+      "step": 2962
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.618398955028923e-06,
+      "loss": 0.1575,
+      "step": 2963
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.618265667901794e-06,
+      "loss": 0.1075,
+      "step": 2964
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.618132380774665e-06,
+      "loss": 0.0909,
+      "step": 2965
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.617999093647536e-06,
+      "loss": 0.1409,
+      "step": 2966
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.617865806520407e-06,
+      "loss": 0.1043,
+      "step": 2967
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.617732519393278e-06,
+      "loss": 0.14,
+      "step": 2968
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.617599232266148e-06,
+      "loss": 0.0988,
+      "step": 2969
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.617465945139019e-06,
+      "loss": 0.1071,
+      "step": 2970
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.61733265801189e-06,
+      "loss": 0.0905,
+      "step": 2971
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.61719937088476e-06,
+      "loss": 0.1128,
+      "step": 2972
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.617066083757632e-06,
+      "loss": 0.1511,
+      "step": 2973
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.616932796630503e-06,
+      "loss": 0.1291,
+      "step": 2974
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.616799509503372e-06,
+      "loss": 0.1051,
+      "step": 2975
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.616666222376245e-06,
+      "loss": 0.1081,
+      "step": 2976
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.616532935249114e-06,
+      "loss": 0.1059,
+      "step": 2977
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.616399648121985e-06,
+      "loss": 0.1569,
+      "step": 2978
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.616266360994856e-06,
+      "loss": 0.0989,
+      "step": 2979
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.616133073867725e-06,
+      "loss": 0.113,
+      "step": 2980
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.615999786740598e-06,
+      "loss": 0.0901,
+      "step": 2981
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.615866499613467e-06,
+      "loss": 0.1136,
+      "step": 2982
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.615733212486338e-06,
+      "loss": 0.1386,
+      "step": 2983
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.61559992535921e-06,
+      "loss": 0.1315,
+      "step": 2984
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.61546663823208e-06,
+      "loss": 0.1106,
+      "step": 2985
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.615333351104951e-06,
+      "loss": 0.1306,
+      "step": 2986
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.615200063977822e-06,
+      "loss": 0.1254,
+      "step": 2987
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.615066776850692e-06,
+      "loss": 0.1567,
+      "step": 2988
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.614933489723564e-06,
+      "loss": 0.1245,
+      "step": 2989
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.614800202596434e-06,
+      "loss": 0.1418,
+      "step": 2990
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.614666915469305e-06,
+      "loss": 0.1258,
+      "step": 2991
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.614533628342176e-06,
+      "loss": 0.1067,
+      "step": 2992
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.614400341215047e-06,
+      "loss": 0.106,
+      "step": 2993
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.614267054087917e-06,
+      "loss": 0.0893,
+      "step": 2994
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.614133766960788e-06,
+      "loss": 0.106,
+      "step": 2995
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.614000479833658e-06,
+      "loss": 0.1508,
+      "step": 2996
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.61386719270653e-06,
+      "loss": 0.1488,
+      "step": 2997
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.6137339055794e-06,
+      "loss": 0.1498,
+      "step": 2998
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.61360061845227e-06,
+      "loss": 0.1165,
+      "step": 2999
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.613467331325142e-06,
+      "loss": 0.1083,
+      "step": 3000
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.613334044198011e-06,
+      "loss": 0.1203,
+      "step": 3001
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.613200757070884e-06,
+      "loss": 0.1534,
+      "step": 3002
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.613067469943753e-06,
+      "loss": 0.0943,
+      "step": 3003
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.612934182816624e-06,
+      "loss": 0.1143,
+      "step": 3004
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.612800895689495e-06,
+      "loss": 0.1242,
+      "step": 3005
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.612667608562366e-06,
+      "loss": 0.0931,
+      "step": 3006
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.612534321435237e-06,
+      "loss": 0.1048,
+      "step": 3007
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.612401034308108e-06,
+      "loss": 0.1184,
+      "step": 3008
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.612267747180977e-06,
+      "loss": 0.1318,
+      "step": 3009
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.612134460053848e-06,
+      "loss": 0.1338,
+      "step": 3010
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.61200117292672e-06,
+      "loss": 0.1125,
+      "step": 3011
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.61186788579959e-06,
+      "loss": 0.1178,
+      "step": 3012
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.611734598672461e-06,
+      "loss": 0.0959,
+      "step": 3013
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.611601311545332e-06,
+      "loss": 0.1114,
+      "step": 3014
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.611468024418202e-06,
+      "loss": 0.1396,
+      "step": 3015
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.611334737291074e-06,
+      "loss": 0.1316,
+      "step": 3016
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.611201450163944e-06,
+      "loss": 0.1005,
+      "step": 3017
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.611068163036815e-06,
+      "loss": 0.1285,
+      "step": 3018
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.610934875909686e-06,
+      "loss": 0.1133,
+      "step": 3019
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.610801588782555e-06,
+      "loss": 0.1083,
+      "step": 3020
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.610668301655428e-06,
+      "loss": 0.1527,
+      "step": 3021
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.610535014528297e-06,
+      "loss": 0.0856,
+      "step": 3022
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.610401727401168e-06,
+      "loss": 0.1393,
+      "step": 3023
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.610268440274039e-06,
+      "loss": 0.1366,
+      "step": 3024
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.61013515314691e-06,
+      "loss": 0.0786,
+      "step": 3025
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.610001866019781e-06,
+      "loss": 0.119,
+      "step": 3026
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.609868578892652e-06,
+      "loss": 0.1305,
+      "step": 3027
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.609735291765521e-06,
+      "loss": 0.1516,
+      "step": 3028
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.609602004638394e-06,
+      "loss": 0.1466,
+      "step": 3029
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.609468717511263e-06,
+      "loss": 0.1222,
+      "step": 3030
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.609335430384134e-06,
+      "loss": 0.1237,
+      "step": 3031
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.609202143257005e-06,
+      "loss": 0.1063,
+      "step": 3032
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.609068856129876e-06,
+      "loss": 0.1089,
+      "step": 3033
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.608935569002747e-06,
+      "loss": 0.1411,
+      "step": 3034
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.608802281875618e-06,
+      "loss": 0.1027,
+      "step": 3035
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.608668994748487e-06,
+      "loss": 0.0898,
+      "step": 3036
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.608535707621358e-06,
+      "loss": 0.115,
+      "step": 3037
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.60840242049423e-06,
+      "loss": 0.1595,
+      "step": 3038
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.6082691333671e-06,
+      "loss": 0.1269,
+      "step": 3039
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.608135846239971e-06,
+      "loss": 0.1364,
+      "step": 3040
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.60800255911284e-06,
+      "loss": 0.1365,
+      "step": 3041
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.607869271985712e-06,
+      "loss": 0.1125,
+      "step": 3042
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.607735984858583e-06,
+      "loss": 0.1864,
+      "step": 3043
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.607602697731454e-06,
+      "loss": 0.1531,
+      "step": 3044
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.607469410604325e-06,
+      "loss": 0.1296,
+      "step": 3045
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.607336123477196e-06,
+      "loss": 0.1246,
+      "step": 3046
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.607202836350065e-06,
+      "loss": 0.1237,
+      "step": 3047
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.607069549222938e-06,
+      "loss": 0.1325,
+      "step": 3048
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.606936262095807e-06,
+      "loss": 0.1513,
+      "step": 3049
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.606802974968678e-06,
+      "loss": 0.1269,
+      "step": 3050
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.606669687841549e-06,
+      "loss": 0.1262,
+      "step": 3051
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.60653640071442e-06,
+      "loss": 0.1192,
+      "step": 3052
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.606403113587291e-06,
+      "loss": 0.1227,
+      "step": 3053
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.606269826460162e-06,
+      "loss": 0.1167,
+      "step": 3054
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.606136539333031e-06,
+      "loss": 0.1288,
+      "step": 3055
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.606003252205904e-06,
+      "loss": 0.1412,
+      "step": 3056
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.605869965078773e-06,
+      "loss": 0.1048,
+      "step": 3057
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.605736677951644e-06,
+      "loss": 0.2186,
+      "step": 3058
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.605603390824515e-06,
+      "loss": 0.1774,
+      "step": 3059
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.605470103697385e-06,
+      "loss": 0.1419,
+      "step": 3060
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.605336816570257e-06,
+      "loss": 0.0833,
+      "step": 3061
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.605203529443127e-06,
+      "loss": 0.1468,
+      "step": 3062
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.605070242315997e-06,
+      "loss": 0.1259,
+      "step": 3063
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.604936955188868e-06,
+      "loss": 0.1157,
+      "step": 3064
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.60480366806174e-06,
+      "loss": 0.116,
+      "step": 3065
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.60467038093461e-06,
+      "loss": 0.1213,
+      "step": 3066
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.604537093807481e-06,
+      "loss": 0.132,
+      "step": 3067
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.60440380668035e-06,
+      "loss": 0.1215,
+      "step": 3068
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.604270519553223e-06,
+      "loss": 0.141,
+      "step": 3069
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.604137232426093e-06,
+      "loss": 0.1288,
+      "step": 3070
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.604003945298964e-06,
+      "loss": 0.1604,
+      "step": 3071
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.603870658171835e-06,
+      "loss": 0.1711,
+      "step": 3072
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.603737371044706e-06,
+      "loss": 0.1262,
+      "step": 3073
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.603604083917577e-06,
+      "loss": 0.0948,
+      "step": 3074
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.603470796790448e-06,
+      "loss": 0.1259,
+      "step": 3075
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.603337509663317e-06,
+      "loss": 0.2001,
+      "step": 3076
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.603204222536188e-06,
+      "loss": 0.137,
+      "step": 3077
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.603070935409059e-06,
+      "loss": 0.1197,
+      "step": 3078
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.60293764828193e-06,
+      "loss": 0.0853,
+      "step": 3079
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.602804361154801e-06,
+      "loss": 0.1417,
+      "step": 3080
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.60267107402767e-06,
+      "loss": 0.0949,
+      "step": 3081
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.602537786900541e-06,
+      "loss": 0.1293,
+      "step": 3082
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.602404499773412e-06,
+      "loss": 0.1375,
+      "step": 3083
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.602271212646283e-06,
+      "loss": 0.1027,
+      "step": 3084
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.602137925519154e-06,
+      "loss": 0.1092,
+      "step": 3085
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.602004638392025e-06,
+      "loss": 0.1382,
+      "step": 3086
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.601871351264895e-06,
+      "loss": 0.1086,
+      "step": 3087
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.601738064137767e-06,
+      "loss": 0.1307,
+      "step": 3088
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.601604777010637e-06,
+      "loss": 0.125,
+      "step": 3089
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.601471489883508e-06,
+      "loss": 0.1377,
+      "step": 3090
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.601338202756379e-06,
+      "loss": 0.1252,
+      "step": 3091
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.60120491562925e-06,
+      "loss": 0.1165,
+      "step": 3092
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.60107162850212e-06,
+      "loss": 0.0851,
+      "step": 3093
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.600938341374992e-06,
+      "loss": 0.1233,
+      "step": 3094
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.600805054247861e-06,
+      "loss": 0.1212,
+      "step": 3095
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.600671767120734e-06,
+      "loss": 0.0971,
+      "step": 3096
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.600538479993603e-06,
+      "loss": 0.1253,
+      "step": 3097
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.600405192866474e-06,
+      "loss": 0.1358,
+      "step": 3098
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.600271905739345e-06,
+      "loss": 0.1462,
+      "step": 3099
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.600138618612214e-06,
+      "loss": 0.1179,
+      "step": 3100
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.600005331485087e-06,
+      "loss": 0.1453,
+      "step": 3101
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.599872044357956e-06,
+      "loss": 0.071,
+      "step": 3102
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.599738757230827e-06,
+      "loss": 0.1156,
+      "step": 3103
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.599605470103698e-06,
+      "loss": 0.144,
+      "step": 3104
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.599472182976569e-06,
+      "loss": 0.0823,
+      "step": 3105
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.59933889584944e-06,
+      "loss": 0.1396,
+      "step": 3106
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.599205608722311e-06,
+      "loss": 0.1883,
+      "step": 3107
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.59907232159518e-06,
+      "loss": 0.0852,
+      "step": 3108
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.598939034468051e-06,
+      "loss": 0.1282,
+      "step": 3109
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.598805747340922e-06,
+      "loss": 0.1114,
+      "step": 3110
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.598672460213793e-06,
+      "loss": 0.1009,
+      "step": 3111
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.598539173086664e-06,
+      "loss": 0.1226,
+      "step": 3112
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.598405885959535e-06,
+      "loss": 0.112,
+      "step": 3113
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.598272598832405e-06,
+      "loss": 0.1071,
+      "step": 3114
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.598139311705277e-06,
+      "loss": 0.1568,
+      "step": 3115
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.598006024578147e-06,
+      "loss": 0.1239,
+      "step": 3116
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.597872737451018e-06,
+      "loss": 0.1292,
+      "step": 3117
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.597739450323889e-06,
+      "loss": 0.1417,
+      "step": 3118
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.597606163196758e-06,
+      "loss": 0.0819,
+      "step": 3119
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.59747287606963e-06,
+      "loss": 0.1207,
+      "step": 3120
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.5973395889425e-06,
+      "loss": 0.0945,
+      "step": 3121
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.597206301815371e-06,
+      "loss": 0.0917,
+      "step": 3122
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.597073014688242e-06,
+      "loss": 0.1218,
+      "step": 3123
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.596939727561113e-06,
+      "loss": 0.1144,
+      "step": 3124
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.596806440433984e-06,
+      "loss": 0.11,
+      "step": 3125
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.596673153306855e-06,
+      "loss": 0.1114,
+      "step": 3126
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.596539866179724e-06,
+      "loss": 0.0925,
+      "step": 3127
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.596406579052597e-06,
+      "loss": 0.1331,
+      "step": 3128
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.596273291925466e-06,
+      "loss": 0.111,
+      "step": 3129
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.596140004798337e-06,
+      "loss": 0.1327,
+      "step": 3130
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.596006717671208e-06,
+      "loss": 0.1006,
+      "step": 3131
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.59587343054408e-06,
+      "loss": 0.1264,
+      "step": 3132
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.59574014341695e-06,
+      "loss": 0.1344,
+      "step": 3133
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.595606856289821e-06,
+      "loss": 0.1162,
+      "step": 3134
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.59547356916269e-06,
+      "loss": 0.115,
+      "step": 3135
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.595340282035563e-06,
+      "loss": 0.1342,
+      "step": 3136
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.595206994908432e-06,
+      "loss": 0.1047,
+      "step": 3137
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.595073707781303e-06,
+      "loss": 0.0804,
+      "step": 3138
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.594940420654174e-06,
+      "loss": 0.158,
+      "step": 3139
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.594807133527044e-06,
+      "loss": 0.1056,
+      "step": 3140
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.594673846399916e-06,
+      "loss": 0.1261,
+      "step": 3141
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.594540559272786e-06,
+      "loss": 0.1305,
+      "step": 3142
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.594407272145657e-06,
+      "loss": 0.1316,
+      "step": 3143
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.594273985018528e-06,
+      "loss": 0.1095,
+      "step": 3144
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.594140697891399e-06,
+      "loss": 0.1728,
+      "step": 3145
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.59400741076427e-06,
+      "loss": 0.1505,
+      "step": 3146
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.59387412363714e-06,
+      "loss": 0.0781,
+      "step": 3147
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.59374083651001e-06,
+      "loss": 0.0959,
+      "step": 3148
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.593607549382881e-06,
+      "loss": 0.083,
+      "step": 3149
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.593474262255752e-06,
+      "loss": 0.1099,
+      "step": 3150
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.593340975128623e-06,
+      "loss": 0.1223,
+      "step": 3151
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.593207688001494e-06,
+      "loss": 0.1118,
+      "step": 3152
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.593074400874365e-06,
+      "loss": 0.1136,
+      "step": 3153
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.592941113747234e-06,
+      "loss": 0.1015,
+      "step": 3154
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.592807826620107e-06,
+      "loss": 0.1588,
+      "step": 3155
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.592674539492976e-06,
+      "loss": 0.092,
+      "step": 3156
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.592541252365847e-06,
+      "loss": 0.0999,
+      "step": 3157
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.592407965238718e-06,
+      "loss": 0.0949,
+      "step": 3158
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.592274678111588e-06,
+      "loss": 0.162,
+      "step": 3159
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.59214139098446e-06,
+      "loss": 0.1565,
+      "step": 3160
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.59200810385733e-06,
+      "loss": 0.1097,
+      "step": 3161
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.5918748167302e-06,
+      "loss": 0.1576,
+      "step": 3162
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.591741529603072e-06,
+      "loss": 0.0847,
+      "step": 3163
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.591608242475943e-06,
+      "loss": 0.1309,
+      "step": 3164
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.591474955348814e-06,
+      "loss": 0.1232,
+      "step": 3165
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.591341668221685e-06,
+      "loss": 0.1069,
+      "step": 3166
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.591208381094554e-06,
+      "loss": 0.1074,
+      "step": 3167
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.591075093967427e-06,
+      "loss": 0.161,
+      "step": 3168
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.590941806840296e-06,
+      "loss": 0.1209,
+      "step": 3169
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.590808519713167e-06,
+      "loss": 0.1014,
+      "step": 3170
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.590675232586038e-06,
+      "loss": 0.1141,
+      "step": 3171
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.590541945458909e-06,
+      "loss": 0.113,
+      "step": 3172
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.59040865833178e-06,
+      "loss": 0.1162,
+      "step": 3173
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.59027537120465e-06,
+      "loss": 0.126,
+      "step": 3174
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.59014208407752e-06,
+      "loss": 0.1189,
+      "step": 3175
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.590008796950391e-06,
+      "loss": 0.0945,
+      "step": 3176
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.589875509823262e-06,
+      "loss": 0.1351,
+      "step": 3177
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.589742222696133e-06,
+      "loss": 0.1595,
+      "step": 3178
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.589608935569004e-06,
+      "loss": 0.1201,
+      "step": 3179
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.589475648441873e-06,
+      "loss": 0.1165,
+      "step": 3180
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.589342361314744e-06,
+      "loss": 0.1419,
+      "step": 3181
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.589209074187615e-06,
+      "loss": 0.1097,
+      "step": 3182
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.589075787060486e-06,
+      "loss": 0.138,
+      "step": 3183
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.588942499933357e-06,
+      "loss": 0.1359,
+      "step": 3184
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.588809212806228e-06,
+      "loss": 0.0932,
+      "step": 3185
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.588675925679098e-06,
+      "loss": 0.1296,
+      "step": 3186
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.58854263855197e-06,
+      "loss": 0.1183,
+      "step": 3187
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.58840935142484e-06,
+      "loss": 0.1137,
+      "step": 3188
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.58827606429771e-06,
+      "loss": 0.159,
+      "step": 3189
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.588142777170582e-06,
+      "loss": 0.119,
+      "step": 3190
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.588009490043453e-06,
+      "loss": 0.1423,
+      "step": 3191
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.587876202916324e-06,
+      "loss": 0.1492,
+      "step": 3192
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.587742915789195e-06,
+      "loss": 0.1336,
+      "step": 3193
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.587609628662064e-06,
+      "loss": 0.1292,
+      "step": 3194
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.587476341534937e-06,
+      "loss": 0.0944,
+      "step": 3195
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.587343054407806e-06,
+      "loss": 0.1143,
+      "step": 3196
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.587209767280677e-06,
+      "loss": 0.1048,
+      "step": 3197
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.587076480153548e-06,
+      "loss": 0.1272,
+      "step": 3198
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.586943193026417e-06,
+      "loss": 0.1092,
+      "step": 3199
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.58680990589929e-06,
+      "loss": 0.1547,
+      "step": 3200
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.58667661877216e-06,
+      "loss": 0.14,
+      "step": 3201
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.58654333164503e-06,
+      "loss": 0.136,
+      "step": 3202
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.586410044517901e-06,
+      "loss": 0.1474,
+      "step": 3203
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.586276757390772e-06,
+      "loss": 0.1278,
+      "step": 3204
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.586143470263643e-06,
+      "loss": 0.1481,
+      "step": 3205
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.586010183136514e-06,
+      "loss": 0.0944,
+      "step": 3206
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.585876896009383e-06,
+      "loss": 0.1084,
+      "step": 3207
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.585743608882254e-06,
+      "loss": 0.0949,
+      "step": 3208
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.585610321755125e-06,
+      "loss": 0.1441,
+      "step": 3209
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.585477034627996e-06,
+      "loss": 0.1097,
+      "step": 3210
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.585343747500867e-06,
+      "loss": 0.1193,
+      "step": 3211
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.585210460373738e-06,
+      "loss": 0.0789,
+      "step": 3212
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.585077173246608e-06,
+      "loss": 0.0884,
+      "step": 3213
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.58494388611948e-06,
+      "loss": 0.1598,
+      "step": 3214
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.58481059899235e-06,
+      "loss": 0.1213,
+      "step": 3215
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.58467731186522e-06,
+      "loss": 0.1396,
+      "step": 3216
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.584544024738092e-06,
+      "loss": 0.0995,
+      "step": 3217
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.584410737610961e-06,
+      "loss": 0.1083,
+      "step": 3218
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.584277450483834e-06,
+      "loss": 0.1546,
+      "step": 3219
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.584144163356703e-06,
+      "loss": 0.1061,
+      "step": 3220
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.584010876229574e-06,
+      "loss": 0.1364,
+      "step": 3221
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.583877589102445e-06,
+      "loss": 0.1339,
+      "step": 3222
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.583744301975316e-06,
+      "loss": 0.0761,
+      "step": 3223
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.583611014848187e-06,
+      "loss": 0.1044,
+      "step": 3224
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.583477727721058e-06,
+      "loss": 0.1406,
+      "step": 3225
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.583344440593927e-06,
+      "loss": 0.1244,
+      "step": 3226
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.5832111534668e-06,
+      "loss": 0.132,
+      "step": 3227
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.58307786633967e-06,
+      "loss": 0.1375,
+      "step": 3228
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.58294457921254e-06,
+      "loss": 0.1404,
+      "step": 3229
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.582811292085411e-06,
+      "loss": 0.0899,
+      "step": 3230
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.582678004958282e-06,
+      "loss": 0.0911,
+      "step": 3231
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.582544717831153e-06,
+      "loss": 0.1315,
+      "step": 3232
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.582411430704024e-06,
+      "loss": 0.1287,
+      "step": 3233
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.582278143576894e-06,
+      "loss": 0.1231,
+      "step": 3234
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.582144856449765e-06,
+      "loss": 0.1144,
+      "step": 3235
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.582011569322636e-06,
+      "loss": 0.1521,
+      "step": 3236
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.581878282195507e-06,
+      "loss": 0.1222,
+      "step": 3237
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.581744995068378e-06,
+      "loss": 0.121,
+      "step": 3238
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.581611707941247e-06,
+      "loss": 0.1494,
+      "step": 3239
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.58147842081412e-06,
+      "loss": 0.114,
+      "step": 3240
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.581345133686989e-06,
+      "loss": 0.1639,
+      "step": 3241
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.58121184655986e-06,
+      "loss": 0.1773,
+      "step": 3242
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.58107855943273e-06,
+      "loss": 0.1299,
+      "step": 3243
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.580945272305602e-06,
+      "loss": 0.112,
+      "step": 3244
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.580811985178473e-06,
+      "loss": 0.1155,
+      "step": 3245
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.580678698051344e-06,
+      "loss": 0.1478,
+      "step": 3246
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.580545410924213e-06,
+      "loss": 0.133,
+      "step": 3247
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.580412123797084e-06,
+      "loss": 0.1227,
+      "step": 3248
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.580278836669955e-06,
+      "loss": 0.0883,
+      "step": 3249
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.580145549542826e-06,
+      "loss": 0.0899,
+      "step": 3250
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.580012262415697e-06,
+      "loss": 0.1378,
+      "step": 3251
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.579878975288568e-06,
+      "loss": 0.1023,
+      "step": 3252
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.579745688161437e-06,
+      "loss": 0.1561,
+      "step": 3253
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.579612401034308e-06,
+      "loss": 0.1317,
+      "step": 3254
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.57947911390718e-06,
+      "loss": 0.0876,
+      "step": 3255
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.57934582678005e-06,
+      "loss": 0.0901,
+      "step": 3256
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.579212539652921e-06,
+      "loss": 0.1477,
+      "step": 3257
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.57907925252579e-06,
+      "loss": 0.0944,
+      "step": 3258
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.578945965398663e-06,
+      "loss": 0.1576,
+      "step": 3259
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.578812678271533e-06,
+      "loss": 0.1283,
+      "step": 3260
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.578679391144404e-06,
+      "loss": 0.0841,
+      "step": 3261
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.578546104017275e-06,
+      "loss": 0.0932,
+      "step": 3262
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.578412816890146e-06,
+      "loss": 0.1077,
+      "step": 3263
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.578279529763017e-06,
+      "loss": 0.0972,
+      "step": 3264
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.578146242635888e-06,
+      "loss": 0.1578,
+      "step": 3265
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.578012955508757e-06,
+      "loss": 0.1276,
+      "step": 3266
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.57787966838163e-06,
+      "loss": 0.1411,
+      "step": 3267
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.577746381254499e-06,
+      "loss": 0.1246,
+      "step": 3268
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.57761309412737e-06,
+      "loss": 0.1105,
+      "step": 3269
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.577479807000241e-06,
+      "loss": 0.1529,
+      "step": 3270
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.577346519873112e-06,
+      "loss": 0.1387,
+      "step": 3271
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.577213232745983e-06,
+      "loss": 0.14,
+      "step": 3272
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.577079945618854e-06,
+      "loss": 0.0896,
+      "step": 3273
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.576946658491723e-06,
+      "loss": 0.0796,
+      "step": 3274
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.576813371364594e-06,
+      "loss": 0.128,
+      "step": 3275
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.576680084237465e-06,
+      "loss": 0.1151,
+      "step": 3276
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.576546797110336e-06,
+      "loss": 0.1204,
+      "step": 3277
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.576413509983207e-06,
+      "loss": 0.1243,
+      "step": 3278
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.576280222856076e-06,
+      "loss": 0.1025,
+      "step": 3279
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.576146935728947e-06,
+      "loss": 0.1278,
+      "step": 3280
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.576013648601818e-06,
+      "loss": 0.1044,
+      "step": 3281
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.57588036147469e-06,
+      "loss": 0.098,
+      "step": 3282
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.57574707434756e-06,
+      "loss": 0.1508,
+      "step": 3283
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.575613787220431e-06,
+      "loss": 0.0936,
+      "step": 3284
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.5754805000933e-06,
+      "loss": 0.0914,
+      "step": 3285
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.575347212966173e-06,
+      "loss": 0.1391,
+      "step": 3286
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.575213925839043e-06,
+      "loss": 0.1249,
+      "step": 3287
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.575080638711914e-06,
+      "loss": 0.1059,
+      "step": 3288
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.574947351584785e-06,
+      "loss": 0.1421,
+      "step": 3289
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.574814064457656e-06,
+      "loss": 0.1091,
+      "step": 3290
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.574680777330527e-06,
+      "loss": 0.0978,
+      "step": 3291
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.574547490203398e-06,
+      "loss": 0.1403,
+      "step": 3292
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.574414203076267e-06,
+      "loss": 0.1477,
+      "step": 3293
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.574280915949138e-06,
+      "loss": 0.1105,
+      "step": 3294
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.574147628822009e-06,
+      "loss": 0.0974,
+      "step": 3295
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.57401434169488e-06,
+      "loss": 0.1285,
+      "step": 3296
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.573881054567751e-06,
+      "loss": 0.1845,
+      "step": 3297
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.57374776744062e-06,
+      "loss": 0.0995,
+      "step": 3298
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.573614480313493e-06,
+      "loss": 0.0964,
+      "step": 3299
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.573481193186362e-06,
+      "loss": 0.1108,
+      "step": 3300
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.573347906059233e-06,
+      "loss": 0.1316,
+      "step": 3301
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.573214618932104e-06,
+      "loss": 0.1581,
+      "step": 3302
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.573081331804975e-06,
+      "loss": 0.106,
+      "step": 3303
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.572948044677846e-06,
+      "loss": 0.1335,
+      "step": 3304
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.572814757550717e-06,
+      "loss": 0.1715,
+      "step": 3305
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.572681470423587e-06,
+      "loss": 0.0995,
+      "step": 3306
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.57254818329646e-06,
+      "loss": 0.1052,
+      "step": 3307
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.572414896169329e-06,
+      "loss": 0.116,
+      "step": 3308
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.5722816090422e-06,
+      "loss": 0.1636,
+      "step": 3309
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.57214832191507e-06,
+      "loss": 0.1387,
+      "step": 3310
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.572015034787942e-06,
+      "loss": 0.1098,
+      "step": 3311
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.571881747660812e-06,
+      "loss": 0.1218,
+      "step": 3312
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.571748460533682e-06,
+      "loss": 0.1551,
+      "step": 3313
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.571615173406553e-06,
+      "loss": 0.1087,
+      "step": 3314
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.571481886279424e-06,
+      "loss": 0.1239,
+      "step": 3315
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.571348599152295e-06,
+      "loss": 0.1096,
+      "step": 3316
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.571215312025166e-06,
+      "loss": 0.1029,
+      "step": 3317
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.571082024898037e-06,
+      "loss": 0.1928,
+      "step": 3318
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.570948737770906e-06,
+      "loss": 0.0942,
+      "step": 3319
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.570815450643777e-06,
+      "loss": 0.122,
+      "step": 3320
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.570682163516648e-06,
+      "loss": 0.1329,
+      "step": 3321
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.570548876389519e-06,
+      "loss": 0.0859,
+      "step": 3322
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.57041558926239e-06,
+      "loss": 0.1124,
+      "step": 3323
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.570282302135261e-06,
+      "loss": 0.1249,
+      "step": 3324
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.57014901500813e-06,
+      "loss": 0.1109,
+      "step": 3325
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.570015727881003e-06,
+      "loss": 0.1562,
+      "step": 3326
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.569882440753872e-06,
+      "loss": 0.1543,
+      "step": 3327
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.569749153626743e-06,
+      "loss": 0.1373,
+      "step": 3328
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.569615866499614e-06,
+      "loss": 0.114,
+      "step": 3329
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.569482579372485e-06,
+      "loss": 0.0981,
+      "step": 3330
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.569349292245356e-06,
+      "loss": 0.1815,
+      "step": 3331
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.569216005118227e-06,
+      "loss": 0.1298,
+      "step": 3332
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.569082717991097e-06,
+      "loss": 0.1031,
+      "step": 3333
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.568949430863968e-06,
+      "loss": 0.1129,
+      "step": 3334
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.568816143736839e-06,
+      "loss": 0.1374,
+      "step": 3335
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.56868285660971e-06,
+      "loss": 0.1359,
+      "step": 3336
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.56854956948258e-06,
+      "loss": 0.0809,
+      "step": 3337
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.56841628235545e-06,
+      "loss": 0.1457,
+      "step": 3338
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.568282995228323e-06,
+      "loss": 0.205,
+      "step": 3339
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.568149708101192e-06,
+      "loss": 0.1195,
+      "step": 3340
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.568016420974063e-06,
+      "loss": 0.1155,
+      "step": 3341
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.567883133846934e-06,
+      "loss": 0.0997,
+      "step": 3342
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.567749846719805e-06,
+      "loss": 0.1215,
+      "step": 3343
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.567616559592676e-06,
+      "loss": 0.1302,
+      "step": 3344
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.567483272465547e-06,
+      "loss": 0.119,
+      "step": 3345
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.567349985338416e-06,
+      "loss": 0.1228,
+      "step": 3346
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.567216698211287e-06,
+      "loss": 0.1076,
+      "step": 3347
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.567083411084158e-06,
+      "loss": 0.1109,
+      "step": 3348
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.566950123957029e-06,
+      "loss": 0.1361,
+      "step": 3349
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.5668168368299e-06,
+      "loss": 0.1096,
+      "step": 3350
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.566683549702771e-06,
+      "loss": 0.1321,
+      "step": 3351
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.56655026257564e-06,
+      "loss": 0.1325,
+      "step": 3352
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.566416975448511e-06,
+      "loss": 0.1519,
+      "step": 3353
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.566283688321382e-06,
+      "loss": 0.1128,
+      "step": 3354
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.566150401194253e-06,
+      "loss": 0.1028,
+      "step": 3355
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.566017114067124e-06,
+      "loss": 0.1129,
+      "step": 3356
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.565883826939994e-06,
+      "loss": 0.1196,
+      "step": 3357
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.565750539812866e-06,
+      "loss": 0.0843,
+      "step": 3358
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.565617252685736e-06,
+      "loss": 0.1031,
+      "step": 3359
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.565483965558607e-06,
+      "loss": 0.1272,
+      "step": 3360
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.565350678431478e-06,
+      "loss": 0.0957,
+      "step": 3361
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.565217391304349e-06,
+      "loss": 0.126,
+      "step": 3362
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.56508410417722e-06,
+      "loss": 0.1198,
+      "step": 3363
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.56495081705009e-06,
+      "loss": 0.107,
+      "step": 3364
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.56481752992296e-06,
+      "loss": 0.1076,
+      "step": 3365
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.564684242795833e-06,
+      "loss": 0.1115,
+      "step": 3366
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.564550955668702e-06,
+      "loss": 0.0693,
+      "step": 3367
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.564417668541573e-06,
+      "loss": 0.0987,
+      "step": 3368
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.564284381414444e-06,
+      "loss": 0.103,
+      "step": 3369
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.564151094287315e-06,
+      "loss": 0.0829,
+      "step": 3370
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.564017807160186e-06,
+      "loss": 0.1154,
+      "step": 3371
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.563884520033055e-06,
+      "loss": 0.0881,
+      "step": 3372
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.563751232905926e-06,
+      "loss": 0.0818,
+      "step": 3373
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.563617945778797e-06,
+      "loss": 0.1105,
+      "step": 3374
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.563484658651668e-06,
+      "loss": 0.1506,
+      "step": 3375
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.56335137152454e-06,
+      "loss": 0.0936,
+      "step": 3376
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.56321808439741e-06,
+      "loss": 0.1242,
+      "step": 3377
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.56308479727028e-06,
+      "loss": 0.0746,
+      "step": 3378
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.562951510143152e-06,
+      "loss": 0.1118,
+      "step": 3379
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.562818223016022e-06,
+      "loss": 0.1135,
+      "step": 3380
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.562684935888892e-06,
+      "loss": 0.1369,
+      "step": 3381
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.562551648761763e-06,
+      "loss": 0.1274,
+      "step": 3382
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.562418361634634e-06,
+      "loss": 0.0969,
+      "step": 3383
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.562285074507505e-06,
+      "loss": 0.1319,
+      "step": 3384
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.562151787380376e-06,
+      "loss": 0.1483,
+      "step": 3385
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.562018500253246e-06,
+      "loss": 0.1308,
+      "step": 3386
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.561885213126117e-06,
+      "loss": 0.1108,
+      "step": 3387
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.561751925998988e-06,
+      "loss": 0.159,
+      "step": 3388
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.561618638871859e-06,
+      "loss": 0.1395,
+      "step": 3389
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.56148535174473e-06,
+      "loss": 0.1236,
+      "step": 3390
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.561352064617599e-06,
+      "loss": 0.1052,
+      "step": 3391
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.56121877749047e-06,
+      "loss": 0.1264,
+      "step": 3392
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.561085490363341e-06,
+      "loss": 0.1364,
+      "step": 3393
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.560952203236212e-06,
+      "loss": 0.1027,
+      "step": 3394
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.560818916109083e-06,
+      "loss": 0.119,
+      "step": 3395
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.560685628981954e-06,
+      "loss": 0.1289,
+      "step": 3396
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.560552341854823e-06,
+      "loss": 0.1719,
+      "step": 3397
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.560419054727696e-06,
+      "loss": 0.1393,
+      "step": 3398
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.560285767600565e-06,
+      "loss": 0.1077,
+      "step": 3399
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.560152480473436e-06,
+      "loss": 0.123,
+      "step": 3400
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.560019193346307e-06,
+      "loss": 0.1228,
+      "step": 3401
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.559885906219178e-06,
+      "loss": 0.1678,
+      "step": 3402
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.55975261909205e-06,
+      "loss": 0.0833,
+      "step": 3403
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.55961933196492e-06,
+      "loss": 0.1324,
+      "step": 3404
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.55948604483779e-06,
+      "loss": 0.1406,
+      "step": 3405
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.559352757710662e-06,
+      "loss": 0.1374,
+      "step": 3406
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.559219470583532e-06,
+      "loss": 0.1077,
+      "step": 3407
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.559086183456403e-06,
+      "loss": 0.1049,
+      "step": 3408
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.558952896329274e-06,
+      "loss": 0.0846,
+      "step": 3409
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.558819609202145e-06,
+      "loss": 0.1242,
+      "step": 3410
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.558686322075016e-06,
+      "loss": 0.1466,
+      "step": 3411
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.558553034947885e-06,
+      "loss": 0.1413,
+      "step": 3412
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.558419747820756e-06,
+      "loss": 0.1301,
+      "step": 3413
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.558286460693627e-06,
+      "loss": 0.149,
+      "step": 3414
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.558153173566498e-06,
+      "loss": 0.0957,
+      "step": 3415
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.558019886439369e-06,
+      "loss": 0.1076,
+      "step": 3416
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.55788659931224e-06,
+      "loss": 0.1114,
+      "step": 3417
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.557753312185109e-06,
+      "loss": 0.1089,
+      "step": 3418
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.55762002505798e-06,
+      "loss": 0.1195,
+      "step": 3419
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.557486737930851e-06,
+      "loss": 0.1225,
+      "step": 3420
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.557353450803722e-06,
+      "loss": 0.1582,
+      "step": 3421
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.557220163676593e-06,
+      "loss": 0.1161,
+      "step": 3422
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.557086876549464e-06,
+      "loss": 0.1049,
+      "step": 3423
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.556953589422333e-06,
+      "loss": 0.0862,
+      "step": 3424
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.556820302295206e-06,
+      "loss": 0.1207,
+      "step": 3425
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.556687015168075e-06,
+      "loss": 0.1314,
+      "step": 3426
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.556553728040946e-06,
+      "loss": 0.1019,
+      "step": 3427
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.556420440913817e-06,
+      "loss": 0.1045,
+      "step": 3428
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.556287153786688e-06,
+      "loss": 0.0889,
+      "step": 3429
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.55615386665956e-06,
+      "loss": 0.1302,
+      "step": 3430
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.556020579532429e-06,
+      "loss": 0.0885,
+      "step": 3431
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.5558872924053e-06,
+      "loss": 0.148,
+      "step": 3432
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.55575400527817e-06,
+      "loss": 0.0905,
+      "step": 3433
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.555620718151042e-06,
+      "loss": 0.1386,
+      "step": 3434
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.555487431023913e-06,
+      "loss": 0.119,
+      "step": 3435
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.555354143896784e-06,
+      "loss": 0.1317,
+      "step": 3436
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.555220856769653e-06,
+      "loss": 0.1021,
+      "step": 3437
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.555087569642526e-06,
+      "loss": 0.1395,
+      "step": 3438
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.554954282515395e-06,
+      "loss": 0.136,
+      "step": 3439
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.554820995388266e-06,
+      "loss": 0.1418,
+      "step": 3440
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.554687708261137e-06,
+      "loss": 0.1299,
+      "step": 3441
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.554554421134008e-06,
+      "loss": 0.099,
+      "step": 3442
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.554421134006879e-06,
+      "loss": 0.1499,
+      "step": 3443
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.55428784687975e-06,
+      "loss": 0.1902,
+      "step": 3444
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.55415455975262e-06,
+      "loss": 0.1596,
+      "step": 3445
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.55402127262549e-06,
+      "loss": 0.1703,
+      "step": 3446
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.553887985498361e-06,
+      "loss": 0.1062,
+      "step": 3447
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.553754698371232e-06,
+      "loss": 0.0932,
+      "step": 3448
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.553621411244103e-06,
+      "loss": 0.1343,
+      "step": 3449
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.553488124116972e-06,
+      "loss": 0.102,
+      "step": 3450
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.553354836989843e-06,
+      "loss": 0.1569,
+      "step": 3451
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.553221549862714e-06,
+      "loss": 0.125,
+      "step": 3452
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.553088262735585e-06,
+      "loss": 0.1111,
+      "step": 3453
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.552954975608456e-06,
+      "loss": 0.1274,
+      "step": 3454
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.552821688481327e-06,
+      "loss": 0.1713,
+      "step": 3455
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.552688401354197e-06,
+      "loss": 0.1163,
+      "step": 3456
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.55255511422707e-06,
+      "loss": 0.0991,
+      "step": 3457
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.552421827099939e-06,
+      "loss": 0.1229,
+      "step": 3458
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.55228853997281e-06,
+      "loss": 0.1104,
+      "step": 3459
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.55215525284568e-06,
+      "loss": 0.1247,
+      "step": 3460
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.552021965718552e-06,
+      "loss": 0.1656,
+      "step": 3461
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.551888678591423e-06,
+      "loss": 0.145,
+      "step": 3462
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.551755391464294e-06,
+      "loss": 0.1054,
+      "step": 3463
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.551622104337163e-06,
+      "loss": 0.1337,
+      "step": 3464
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.551488817210036e-06,
+      "loss": 0.1196,
+      "step": 3465
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.551355530082905e-06,
+      "loss": 0.111,
+      "step": 3466
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.551222242955776e-06,
+      "loss": 0.0944,
+      "step": 3467
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.551088955828647e-06,
+      "loss": 0.1171,
+      "step": 3468
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.550955668701518e-06,
+      "loss": 0.139,
+      "step": 3469
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.550822381574389e-06,
+      "loss": 0.1224,
+      "step": 3470
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.550689094447258e-06,
+      "loss": 0.1073,
+      "step": 3471
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.55055580732013e-06,
+      "loss": 0.1339,
+      "step": 3472
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.550422520193e-06,
+      "loss": 0.1727,
+      "step": 3473
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.550289233065871e-06,
+      "loss": 0.1168,
+      "step": 3474
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.550155945938742e-06,
+      "loss": 0.0937,
+      "step": 3475
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.550022658811613e-06,
+      "loss": 0.143,
+      "step": 3476
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.549889371684483e-06,
+      "loss": 0.1161,
+      "step": 3477
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.549756084557355e-06,
+      "loss": 0.0919,
+      "step": 3478
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.549622797430225e-06,
+      "loss": 0.1326,
+      "step": 3479
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.549489510303096e-06,
+      "loss": 0.1221,
+      "step": 3480
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.549356223175967e-06,
+      "loss": 0.0896,
+      "step": 3481
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.549222936048838e-06,
+      "loss": 0.1164,
+      "step": 3482
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.549089648921709e-06,
+      "loss": 0.1141,
+      "step": 3483
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.54895636179458e-06,
+      "loss": 0.1639,
+      "step": 3484
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.548823074667449e-06,
+      "loss": 0.1067,
+      "step": 3485
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.54868978754032e-06,
+      "loss": 0.1136,
+      "step": 3486
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.54855650041319e-06,
+      "loss": 0.0973,
+      "step": 3487
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.548423213286062e-06,
+      "loss": 0.1385,
+      "step": 3488
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.548289926158933e-06,
+      "loss": 0.1368,
+      "step": 3489
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.548156639031802e-06,
+      "loss": 0.1297,
+      "step": 3490
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.548023351904673e-06,
+      "loss": 0.138,
+      "step": 3491
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.547890064777544e-06,
+      "loss": 0.141,
+      "step": 3492
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.547756777650415e-06,
+      "loss": 0.1393,
+      "step": 3493
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.547623490523286e-06,
+      "loss": 0.1035,
+      "step": 3494
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.547490203396157e-06,
+      "loss": 0.0807,
+      "step": 3495
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.547356916269026e-06,
+      "loss": 0.1191,
+      "step": 3496
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.547223629141899e-06,
+      "loss": 0.1483,
+      "step": 3497
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.547090342014768e-06,
+      "loss": 0.1417,
+      "step": 3498
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.54695705488764e-06,
+      "loss": 0.1014,
+      "step": 3499
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.54682376776051e-06,
+      "loss": 0.1249,
+      "step": 3500
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.546690480633381e-06,
+      "loss": 0.1079,
+      "step": 3501
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.546557193506252e-06,
+      "loss": 0.1256,
+      "step": 3502
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.546423906379123e-06,
+      "loss": 0.0883,
+      "step": 3503
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.546290619251993e-06,
+      "loss": 0.1061,
+      "step": 3504
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.546157332124865e-06,
+      "loss": 0.1117,
+      "step": 3505
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.546024044997735e-06,
+      "loss": 0.1534,
+      "step": 3506
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.545890757870606e-06,
+      "loss": 0.1148,
+      "step": 3507
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.545757470743477e-06,
+      "loss": 0.0987,
+      "step": 3508
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.545624183616346e-06,
+      "loss": 0.1302,
+      "step": 3509
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.545490896489219e-06,
+      "loss": 0.1038,
+      "step": 3510
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.545357609362088e-06,
+      "loss": 0.1427,
+      "step": 3511
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.545224322234959e-06,
+      "loss": 0.0984,
+      "step": 3512
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.54509103510783e-06,
+      "loss": 0.1041,
+      "step": 3513
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.544957747980701e-06,
+      "loss": 0.1279,
+      "step": 3514
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.544824460853572e-06,
+      "loss": 0.078,
+      "step": 3515
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.544691173726443e-06,
+      "loss": 0.1033,
+      "step": 3516
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.544557886599312e-06,
+      "loss": 0.172,
+      "step": 3517
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.544424599472183e-06,
+      "loss": 0.1039,
+      "step": 3518
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.544291312345054e-06,
+      "loss": 0.1082,
+      "step": 3519
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.544158025217925e-06,
+      "loss": 0.1304,
+      "step": 3520
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.544024738090796e-06,
+      "loss": 0.1229,
+      "step": 3521
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.543891450963667e-06,
+      "loss": 0.1037,
+      "step": 3522
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.543758163836536e-06,
+      "loss": 0.128,
+      "step": 3523
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.54362487670941e-06,
+      "loss": 0.1512,
+      "step": 3524
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.543491589582278e-06,
+      "loss": 0.111,
+      "step": 3525
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.54335830245515e-06,
+      "loss": 0.1196,
+      "step": 3526
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.54322501532802e-06,
+      "loss": 0.1199,
+      "step": 3527
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.543091728200891e-06,
+      "loss": 0.0867,
+      "step": 3528
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.542958441073762e-06,
+      "loss": 0.1213,
+      "step": 3529
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.542825153946632e-06,
+      "loss": 0.1137,
+      "step": 3530
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.542691866819503e-06,
+      "loss": 0.1406,
+      "step": 3531
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.542558579692374e-06,
+      "loss": 0.0914,
+      "step": 3532
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.542425292565245e-06,
+      "loss": 0.1408,
+      "step": 3533
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.542292005438116e-06,
+      "loss": 0.1417,
+      "step": 3534
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.542158718310987e-06,
+      "loss": 0.0962,
+      "step": 3535
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.542025431183856e-06,
+      "loss": 0.154,
+      "step": 3536
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.541892144056729e-06,
+      "loss": 0.1463,
+      "step": 3537
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.541758856929598e-06,
+      "loss": 0.089,
+      "step": 3538
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.541625569802469e-06,
+      "loss": 0.0691,
+      "step": 3539
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.54149228267534e-06,
+      "loss": 0.0567,
+      "step": 3540
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.541358995548211e-06,
+      "loss": 0.1235,
+      "step": 3541
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.541225708421082e-06,
+      "loss": 0.1661,
+      "step": 3542
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.541092421293953e-06,
+      "loss": 0.0914,
+      "step": 3543
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.540959134166822e-06,
+      "loss": 0.1225,
+      "step": 3544
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.540825847039695e-06,
+      "loss": 0.1607,
+      "step": 3545
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.540692559912564e-06,
+      "loss": 0.1341,
+      "step": 3546
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.540559272785435e-06,
+      "loss": 0.1475,
+      "step": 3547
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.540425985658306e-06,
+      "loss": 0.1369,
+      "step": 3548
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.540292698531176e-06,
+      "loss": 0.1402,
+      "step": 3549
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.540159411404048e-06,
+      "loss": 0.1222,
+      "step": 3550
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.540026124276918e-06,
+      "loss": 0.2039,
+      "step": 3551
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.539892837149789e-06,
+      "loss": 0.0672,
+      "step": 3552
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.53975955002266e-06,
+      "loss": 0.1225,
+      "step": 3553
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.53962626289553e-06,
+      "loss": 0.1477,
+      "step": 3554
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.539492975768402e-06,
+      "loss": 0.092,
+      "step": 3555
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.539359688641273e-06,
+      "loss": 0.1416,
+      "step": 3556
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.539226401514142e-06,
+      "loss": 0.1612,
+      "step": 3557
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.539093114387013e-06,
+      "loss": 0.1252,
+      "step": 3558
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.538959827259884e-06,
+      "loss": 0.1134,
+      "step": 3559
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.538826540132755e-06,
+      "loss": 0.1571,
+      "step": 3560
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.538693253005626e-06,
+      "loss": 0.1746,
+      "step": 3561
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.538559965878497e-06,
+      "loss": 0.1206,
+      "step": 3562
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.538426678751366e-06,
+      "loss": 0.1103,
+      "step": 3563
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.538293391624239e-06,
+      "loss": 0.1274,
+      "step": 3564
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.538160104497108e-06,
+      "loss": 0.1239,
+      "step": 3565
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.538026817369979e-06,
+      "loss": 0.1392,
+      "step": 3566
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.53789353024285e-06,
+      "loss": 0.1105,
+      "step": 3567
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.53776024311572e-06,
+      "loss": 0.106,
+      "step": 3568
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.537626955988592e-06,
+      "loss": 0.1476,
+      "step": 3569
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.537493668861461e-06,
+      "loss": 0.0957,
+      "step": 3570
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.537360381734332e-06,
+      "loss": 0.1111,
+      "step": 3571
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.537227094607203e-06,
+      "loss": 0.1153,
+      "step": 3572
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.537093807480074e-06,
+      "loss": 0.1095,
+      "step": 3573
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.536960520352945e-06,
+      "loss": 0.1091,
+      "step": 3574
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.536827233225816e-06,
+      "loss": 0.1485,
+      "step": 3575
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.536693946098686e-06,
+      "loss": 0.1298,
+      "step": 3576
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.536560658971558e-06,
+      "loss": 0.1125,
+      "step": 3577
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.536427371844428e-06,
+      "loss": 0.0743,
+      "step": 3578
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.536294084717299e-06,
+      "loss": 0.1306,
+      "step": 3579
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.53616079759017e-06,
+      "loss": 0.0939,
+      "step": 3580
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.53602751046304e-06,
+      "loss": 0.1505,
+      "step": 3581
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.535894223335912e-06,
+      "loss": 0.1361,
+      "step": 3582
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.535760936208783e-06,
+      "loss": 0.0859,
+      "step": 3583
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.535627649081652e-06,
+      "loss": 0.1323,
+      "step": 3584
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.535494361954523e-06,
+      "loss": 0.1112,
+      "step": 3585
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.535361074827394e-06,
+      "loss": 0.1105,
+      "step": 3586
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.535227787700265e-06,
+      "loss": 0.1031,
+      "step": 3587
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.535094500573136e-06,
+      "loss": 0.1231,
+      "step": 3588
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.534961213446005e-06,
+      "loss": 0.0917,
+      "step": 3589
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.534827926318876e-06,
+      "loss": 0.1189,
+      "step": 3590
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.534694639191747e-06,
+      "loss": 0.1284,
+      "step": 3591
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.534561352064618e-06,
+      "loss": 0.1239,
+      "step": 3592
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.53442806493749e-06,
+      "loss": 0.2135,
+      "step": 3593
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.53429477781036e-06,
+      "loss": 0.0916,
+      "step": 3594
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.53416149068323e-06,
+      "loss": 0.1015,
+      "step": 3595
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.534028203556102e-06,
+      "loss": 0.1774,
+      "step": 3596
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.533894916428971e-06,
+      "loss": 0.104,
+      "step": 3597
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.533761629301842e-06,
+      "loss": 0.1051,
+      "step": 3598
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.533628342174713e-06,
+      "loss": 0.1217,
+      "step": 3599
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.533495055047584e-06,
+      "loss": 0.1386,
+      "step": 3600
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.533361767920455e-06,
+      "loss": 0.0939,
+      "step": 3601
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.533228480793326e-06,
+      "loss": 0.0847,
+      "step": 3602
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.533095193666196e-06,
+      "loss": 0.1035,
+      "step": 3603
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.532961906539068e-06,
+      "loss": 0.1247,
+      "step": 3604
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.532828619411938e-06,
+      "loss": 0.0824,
+      "step": 3605
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.532695332284809e-06,
+      "loss": 0.1209,
+      "step": 3606
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.53256204515768e-06,
+      "loss": 0.1523,
+      "step": 3607
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.532428758030549e-06,
+      "loss": 0.1493,
+      "step": 3608
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.532295470903422e-06,
+      "loss": 0.0864,
+      "step": 3609
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.532162183776291e-06,
+      "loss": 0.1256,
+      "step": 3610
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.532028896649162e-06,
+      "loss": 0.1208,
+      "step": 3611
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.531895609522033e-06,
+      "loss": 0.1056,
+      "step": 3612
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.531762322394904e-06,
+      "loss": 0.0933,
+      "step": 3613
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.531629035267775e-06,
+      "loss": 0.1012,
+      "step": 3614
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.531495748140646e-06,
+      "loss": 0.1096,
+      "step": 3615
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.531362461013515e-06,
+      "loss": 0.0875,
+      "step": 3616
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.531229173886388e-06,
+      "loss": 0.1069,
+      "step": 3617
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.531095886759257e-06,
+      "loss": 0.1023,
+      "step": 3618
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.530962599632128e-06,
+      "loss": 0.132,
+      "step": 3619
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.530829312505e-06,
+      "loss": 0.0902,
+      "step": 3620
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.53069602537787e-06,
+      "loss": 0.0885,
+      "step": 3621
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.530562738250741e-06,
+      "loss": 0.1298,
+      "step": 3622
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.530429451123612e-06,
+      "loss": 0.091,
+      "step": 3623
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.530296163996482e-06,
+      "loss": 0.1368,
+      "step": 3624
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.530162876869353e-06,
+      "loss": 0.1479,
+      "step": 3625
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.530029589742224e-06,
+      "loss": 0.1096,
+      "step": 3626
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.529896302615095e-06,
+      "loss": 0.1031,
+      "step": 3627
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.529763015487966e-06,
+      "loss": 0.082,
+      "step": 3628
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.529629728360835e-06,
+      "loss": 0.1076,
+      "step": 3629
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.529496441233706e-06,
+      "loss": 0.1481,
+      "step": 3630
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.529363154106577e-06,
+      "loss": 0.1302,
+      "step": 3631
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.529229866979448e-06,
+      "loss": 0.1023,
+      "step": 3632
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.529096579852319e-06,
+      "loss": 0.1328,
+      "step": 3633
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.52896329272519e-06,
+      "loss": 0.125,
+      "step": 3634
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.528830005598059e-06,
+      "loss": 0.1498,
+      "step": 3635
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.528696718470932e-06,
+      "loss": 0.1439,
+      "step": 3636
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.528563431343801e-06,
+      "loss": 0.0935,
+      "step": 3637
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.528430144216672e-06,
+      "loss": 0.1265,
+      "step": 3638
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.528296857089543e-06,
+      "loss": 0.0791,
+      "step": 3639
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.528163569962414e-06,
+      "loss": 0.1002,
+      "step": 3640
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.528030282835285e-06,
+      "loss": 0.1422,
+      "step": 3641
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.527896995708156e-06,
+      "loss": 0.1323,
+      "step": 3642
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.527763708581025e-06,
+      "loss": 0.1027,
+      "step": 3643
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.527630421453898e-06,
+      "loss": 0.1082,
+      "step": 3644
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.527497134326767e-06,
+      "loss": 0.118,
+      "step": 3645
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.527363847199638e-06,
+      "loss": 0.132,
+      "step": 3646
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.52723056007251e-06,
+      "loss": 0.0966,
+      "step": 3647
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.527097272945379e-06,
+      "loss": 0.1234,
+      "step": 3648
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.526963985818251e-06,
+      "loss": 0.131,
+      "step": 3649
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.52683069869112e-06,
+      "loss": 0.1144,
+      "step": 3650
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.526697411563992e-06,
+      "loss": 0.1546,
+      "step": 3651
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.526564124436863e-06,
+      "loss": 0.1495,
+      "step": 3652
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.526430837309734e-06,
+      "loss": 0.0921,
+      "step": 3653
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.526297550182605e-06,
+      "loss": 0.1341,
+      "step": 3654
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.526164263055476e-06,
+      "loss": 0.1039,
+      "step": 3655
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.526030975928345e-06,
+      "loss": 0.1147,
+      "step": 3656
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.525897688801216e-06,
+      "loss": 0.0899,
+      "step": 3657
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.525764401674087e-06,
+      "loss": 0.1039,
+      "step": 3658
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.525631114546958e-06,
+      "loss": 0.1228,
+      "step": 3659
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.525497827419829e-06,
+      "loss": 0.1062,
+      "step": 3660
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.5253645402927e-06,
+      "loss": 0.0837,
+      "step": 3661
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.52523125316557e-06,
+      "loss": 0.1274,
+      "step": 3662
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.525097966038442e-06,
+      "loss": 0.1274,
+      "step": 3663
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.524964678911311e-06,
+      "loss": 0.1261,
+      "step": 3664
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.524831391784182e-06,
+      "loss": 0.0804,
+      "step": 3665
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.524698104657053e-06,
+      "loss": 0.0968,
+      "step": 3666
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.524564817529922e-06,
+      "loss": 0.1584,
+      "step": 3667
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.524431530402795e-06,
+      "loss": 0.1142,
+      "step": 3668
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.524298243275664e-06,
+      "loss": 0.1305,
+      "step": 3669
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.524164956148535e-06,
+      "loss": 0.1562,
+      "step": 3670
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.524031669021406e-06,
+      "loss": 0.1163,
+      "step": 3671
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.523898381894277e-06,
+      "loss": 0.1114,
+      "step": 3672
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.523765094767148e-06,
+      "loss": 0.1143,
+      "step": 3673
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.52363180764002e-06,
+      "loss": 0.1158,
+      "step": 3674
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.523498520512889e-06,
+      "loss": 0.1172,
+      "step": 3675
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.523365233385761e-06,
+      "loss": 0.1285,
+      "step": 3676
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.52323194625863e-06,
+      "loss": 0.1377,
+      "step": 3677
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.523098659131502e-06,
+      "loss": 0.1004,
+      "step": 3678
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.522965372004373e-06,
+      "loss": 0.0921,
+      "step": 3679
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.522832084877244e-06,
+      "loss": 0.1054,
+      "step": 3680
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.522698797750115e-06,
+      "loss": 0.1743,
+      "step": 3681
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.522565510622986e-06,
+      "loss": 0.1025,
+      "step": 3682
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.522432223495855e-06,
+      "loss": 0.1234,
+      "step": 3683
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.522298936368728e-06,
+      "loss": 0.1014,
+      "step": 3684
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.522165649241597e-06,
+      "loss": 0.1362,
+      "step": 3685
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.522032362114468e-06,
+      "loss": 0.1021,
+      "step": 3686
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.521899074987339e-06,
+      "loss": 0.1668,
+      "step": 3687
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.521765787860208e-06,
+      "loss": 0.1417,
+      "step": 3688
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.521632500733081e-06,
+      "loss": 0.1624,
+      "step": 3689
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.52149921360595e-06,
+      "loss": 0.138,
+      "step": 3690
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.521365926478821e-06,
+      "loss": 0.1158,
+      "step": 3691
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.521232639351692e-06,
+      "loss": 0.1025,
+      "step": 3692
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.521099352224563e-06,
+      "loss": 0.1138,
+      "step": 3693
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.520966065097434e-06,
+      "loss": 0.1464,
+      "step": 3694
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.520832777970305e-06,
+      "loss": 0.1399,
+      "step": 3695
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.520699490843175e-06,
+      "loss": 0.1165,
+      "step": 3696
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.520566203716046e-06,
+      "loss": 0.0987,
+      "step": 3697
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.520432916588917e-06,
+      "loss": 0.1155,
+      "step": 3698
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.520299629461788e-06,
+      "loss": 0.1019,
+      "step": 3699
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.520166342334658e-06,
+      "loss": 0.0793,
+      "step": 3700
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.52003305520753e-06,
+      "loss": 0.1471,
+      "step": 3701
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.519899768080399e-06,
+      "loss": 0.1224,
+      "step": 3702
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.519766480953271e-06,
+      "loss": 0.0657,
+      "step": 3703
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.51963319382614e-06,
+      "loss": 0.143,
+      "step": 3704
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.519499906699012e-06,
+      "loss": 0.1104,
+      "step": 3705
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.519366619571883e-06,
+      "loss": 0.1437,
+      "step": 3706
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.519233332444752e-06,
+      "loss": 0.1747,
+      "step": 3707
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.519100045317625e-06,
+      "loss": 0.1437,
+      "step": 3708
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.518966758190494e-06,
+      "loss": 0.1514,
+      "step": 3709
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.518833471063365e-06,
+      "loss": 0.1194,
+      "step": 3710
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.518700183936236e-06,
+      "loss": 0.121,
+      "step": 3711
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.518566896809107e-06,
+      "loss": 0.0959,
+      "step": 3712
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.518433609681978e-06,
+      "loss": 0.1371,
+      "step": 3713
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.518300322554849e-06,
+      "loss": 0.111,
+      "step": 3714
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.518167035427718e-06,
+      "loss": 0.0909,
+      "step": 3715
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.518033748300591e-06,
+      "loss": 0.1287,
+      "step": 3716
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.51790046117346e-06,
+      "loss": 0.0917,
+      "step": 3717
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.517767174046331e-06,
+      "loss": 0.1308,
+      "step": 3718
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.517633886919202e-06,
+      "loss": 0.1311,
+      "step": 3719
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.517500599792073e-06,
+      "loss": 0.1327,
+      "step": 3720
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.517367312664944e-06,
+      "loss": 0.1371,
+      "step": 3721
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.517234025537815e-06,
+      "loss": 0.11,
+      "step": 3722
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.517100738410685e-06,
+      "loss": 0.1421,
+      "step": 3723
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.516967451283556e-06,
+      "loss": 0.1925,
+      "step": 3724
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.516834164156427e-06,
+      "loss": 0.1617,
+      "step": 3725
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.516700877029298e-06,
+      "loss": 0.1261,
+      "step": 3726
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.516567589902169e-06,
+      "loss": 0.1043,
+      "step": 3727
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.516434302775038e-06,
+      "loss": 0.1466,
+      "step": 3728
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.516301015647909e-06,
+      "loss": 0.1294,
+      "step": 3729
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.51616772852078e-06,
+      "loss": 0.1059,
+      "step": 3730
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.516034441393651e-06,
+      "loss": 0.0884,
+      "step": 3731
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.515901154266522e-06,
+      "loss": 0.0979,
+      "step": 3732
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.515767867139393e-06,
+      "loss": 0.1256,
+      "step": 3733
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.515634580012262e-06,
+      "loss": 0.1128,
+      "step": 3734
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.515501292885135e-06,
+      "loss": 0.1331,
+      "step": 3735
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.515368005758004e-06,
+      "loss": 0.1233,
+      "step": 3736
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.515234718630875e-06,
+      "loss": 0.1511,
+      "step": 3737
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.515101431503746e-06,
+      "loss": 0.1001,
+      "step": 3738
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.514968144376617e-06,
+      "loss": 0.1286,
+      "step": 3739
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.514834857249488e-06,
+      "loss": 0.1372,
+      "step": 3740
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.514701570122359e-06,
+      "loss": 0.1122,
+      "step": 3741
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.514568282995228e-06,
+      "loss": 0.0951,
+      "step": 3742
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.514434995868101e-06,
+      "loss": 0.1289,
+      "step": 3743
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.51430170874097e-06,
+      "loss": 0.1141,
+      "step": 3744
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.514168421613841e-06,
+      "loss": 0.1185,
+      "step": 3745
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.514035134486712e-06,
+      "loss": 0.1073,
+      "step": 3746
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.513901847359582e-06,
+      "loss": 0.0911,
+      "step": 3747
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.513768560232454e-06,
+      "loss": 0.0983,
+      "step": 3748
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.513635273105324e-06,
+      "loss": 0.0923,
+      "step": 3749
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.513501985978195e-06,
+      "loss": 0.1269,
+      "step": 3750
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.513368698851066e-06,
+      "loss": 0.095,
+      "step": 3751
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.513235411723937e-06,
+      "loss": 0.0954,
+      "step": 3752
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.513102124596808e-06,
+      "loss": 0.1787,
+      "step": 3753
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.512968837469679e-06,
+      "loss": 0.1324,
+      "step": 3754
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.512835550342548e-06,
+      "loss": 0.1024,
+      "step": 3755
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.512702263215419e-06,
+      "loss": 0.1415,
+      "step": 3756
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.51256897608829e-06,
+      "loss": 0.1117,
+      "step": 3757
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.512435688961161e-06,
+      "loss": 0.1486,
+      "step": 3758
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.512302401834032e-06,
+      "loss": 0.1115,
+      "step": 3759
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.512169114706903e-06,
+      "loss": 0.1232,
+      "step": 3760
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.512035827579772e-06,
+      "loss": 0.1327,
+      "step": 3761
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.511902540452645e-06,
+      "loss": 0.1153,
+      "step": 3762
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.511769253325514e-06,
+      "loss": 0.1457,
+      "step": 3763
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.511635966198385e-06,
+      "loss": 0.1461,
+      "step": 3764
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.511502679071256e-06,
+      "loss": 0.0777,
+      "step": 3765
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.511369391944126e-06,
+      "loss": 0.0824,
+      "step": 3766
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.511236104816998e-06,
+      "loss": 0.118,
+      "step": 3767
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.511102817689868e-06,
+      "loss": 0.1425,
+      "step": 3768
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.510969530562738e-06,
+      "loss": 0.1219,
+      "step": 3769
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.51083624343561e-06,
+      "loss": 0.099,
+      "step": 3770
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.51070295630848e-06,
+      "loss": 0.1291,
+      "step": 3771
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.510569669181351e-06,
+      "loss": 0.1318,
+      "step": 3772
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.510436382054222e-06,
+      "loss": 0.0826,
+      "step": 3773
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.510303094927092e-06,
+      "loss": 0.1299,
+      "step": 3774
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.510169807799964e-06,
+      "loss": 0.1008,
+      "step": 3775
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.510036520672834e-06,
+      "loss": 0.1241,
+      "step": 3776
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.509903233545705e-06,
+      "loss": 0.079,
+      "step": 3777
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.509769946418576e-06,
+      "loss": 0.1035,
+      "step": 3778
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.509636659291447e-06,
+      "loss": 0.1319,
+      "step": 3779
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.509503372164318e-06,
+      "loss": 0.1143,
+      "step": 3780
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.509370085037189e-06,
+      "loss": 0.167,
+      "step": 3781
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.509236797910058e-06,
+      "loss": 0.0881,
+      "step": 3782
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.509103510782929e-06,
+      "loss": 0.1372,
+      "step": 3783
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.5089702236558e-06,
+      "loss": 0.1121,
+      "step": 3784
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.508836936528671e-06,
+      "loss": 0.1344,
+      "step": 3785
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.508703649401542e-06,
+      "loss": 0.1581,
+      "step": 3786
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.508570362274411e-06,
+      "loss": 0.1131,
+      "step": 3787
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.508437075147284e-06,
+      "loss": 0.1165,
+      "step": 3788
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.508303788020153e-06,
+      "loss": 0.1312,
+      "step": 3789
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.508170500893024e-06,
+      "loss": 0.1204,
+      "step": 3790
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.508037213765895e-06,
+      "loss": 0.1173,
+      "step": 3791
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.507903926638766e-06,
+      "loss": 0.1023,
+      "step": 3792
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.507770639511637e-06,
+      "loss": 0.1007,
+      "step": 3793
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.507637352384508e-06,
+      "loss": 0.0975,
+      "step": 3794
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.507504065257378e-06,
+      "loss": 0.1068,
+      "step": 3795
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.507370778130249e-06,
+      "loss": 0.1376,
+      "step": 3796
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.50723749100312e-06,
+      "loss": 0.1017,
+      "step": 3797
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.50710420387599e-06,
+      "loss": 0.1787,
+      "step": 3798
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.506970916748862e-06,
+      "loss": 0.1194,
+      "step": 3799
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.506837629621733e-06,
+      "loss": 0.0881,
+      "step": 3800
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.506704342494602e-06,
+      "loss": 0.1108,
+      "step": 3801
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.506571055367475e-06,
+      "loss": 0.1029,
+      "step": 3802
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.506437768240344e-06,
+      "loss": 0.1538,
+      "step": 3803
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.506304481113215e-06,
+      "loss": 0.1236,
+      "step": 3804
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.506171193986086e-06,
+      "loss": 0.1386,
+      "step": 3805
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.506037906858955e-06,
+      "loss": 0.1229,
+      "step": 3806
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.505904619731828e-06,
+      "loss": 0.122,
+      "step": 3807
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.505771332604697e-06,
+      "loss": 0.108,
+      "step": 3808
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.505638045477568e-06,
+      "loss": 0.1086,
+      "step": 3809
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.505504758350439e-06,
+      "loss": 0.1052,
+      "step": 3810
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.50537147122331e-06,
+      "loss": 0.1117,
+      "step": 3811
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.505238184096181e-06,
+      "loss": 0.0724,
+      "step": 3812
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.505104896969052e-06,
+      "loss": 0.1293,
+      "step": 3813
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.504971609841921e-06,
+      "loss": 0.0936,
+      "step": 3814
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.504838322714794e-06,
+      "loss": 0.1421,
+      "step": 3815
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.504705035587663e-06,
+      "loss": 0.1509,
+      "step": 3816
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.504571748460534e-06,
+      "loss": 0.1472,
+      "step": 3817
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.504438461333405e-06,
+      "loss": 0.1722,
+      "step": 3818
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.504305174206276e-06,
+      "loss": 0.1599,
+      "step": 3819
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.504171887079147e-06,
+      "loss": 0.1273,
+      "step": 3820
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.504038599952018e-06,
+      "loss": 0.1369,
+      "step": 3821
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.503905312824888e-06,
+      "loss": 0.0965,
+      "step": 3822
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.503772025697759e-06,
+      "loss": 0.1176,
+      "step": 3823
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.50363873857063e-06,
+      "loss": 0.1351,
+      "step": 3824
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.5035054514435e-06,
+      "loss": 0.0837,
+      "step": 3825
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.503372164316372e-06,
+      "loss": 0.1186,
+      "step": 3826
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.503238877189241e-06,
+      "loss": 0.1187,
+      "step": 3827
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.503105590062112e-06,
+      "loss": 0.1206,
+      "step": 3828
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.502972302934983e-06,
+      "loss": 0.0987,
+      "step": 3829
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.502839015807854e-06,
+      "loss": 0.1133,
+      "step": 3830
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.502705728680725e-06,
+      "loss": 0.1059,
+      "step": 3831
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.502572441553596e-06,
+      "loss": 0.1,
+      "step": 3832
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.502439154426465e-06,
+      "loss": 0.1011,
+      "step": 3833
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.502305867299338e-06,
+      "loss": 0.1001,
+      "step": 3834
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.502172580172207e-06,
+      "loss": 0.075,
+      "step": 3835
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.502039293045078e-06,
+      "loss": 0.1303,
+      "step": 3836
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.50190600591795e-06,
+      "loss": 0.1384,
+      "step": 3837
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.50177271879082e-06,
+      "loss": 0.0961,
+      "step": 3838
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.501639431663691e-06,
+      "loss": 0.1278,
+      "step": 3839
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.501506144536562e-06,
+      "loss": 0.1463,
+      "step": 3840
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.501372857409431e-06,
+      "loss": 0.1162,
+      "step": 3841
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.501239570282302e-06,
+      "loss": 0.0809,
+      "step": 3842
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.501106283155173e-06,
+      "loss": 0.1305,
+      "step": 3843
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.500972996028044e-06,
+      "loss": 0.0905,
+      "step": 3844
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.500839708900915e-06,
+      "loss": 0.1223,
+      "step": 3845
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.500706421773785e-06,
+      "loss": 0.1245,
+      "step": 3846
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.500573134646657e-06,
+      "loss": 0.1205,
+      "step": 3847
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.500439847519527e-06,
+      "loss": 0.119,
+      "step": 3848
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.500306560392398e-06,
+      "loss": 0.0886,
+      "step": 3849
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.500173273265269e-06,
+      "loss": 0.1018,
+      "step": 3850
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.50003998613814e-06,
+      "loss": 0.1404,
+      "step": 3851
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.49990669901101e-06,
+      "loss": 0.1009,
+      "step": 3852
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.499773411883882e-06,
+      "loss": 0.0963,
+      "step": 3853
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.499640124756751e-06,
+      "loss": 0.1296,
+      "step": 3854
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.499506837629624e-06,
+      "loss": 0.1098,
+      "step": 3855
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.499373550502493e-06,
+      "loss": 0.1271,
+      "step": 3856
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.499240263375364e-06,
+      "loss": 0.1117,
+      "step": 3857
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.499106976248235e-06,
+      "loss": 0.113,
+      "step": 3858
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.498973689121106e-06,
+      "loss": 0.0944,
+      "step": 3859
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.498840401993977e-06,
+      "loss": 0.1082,
+      "step": 3860
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.498707114866848e-06,
+      "loss": 0.1781,
+      "step": 3861
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.498573827739717e-06,
+      "loss": 0.1151,
+      "step": 3862
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.498440540612588e-06,
+      "loss": 0.1177,
+      "step": 3863
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.49830725348546e-06,
+      "loss": 0.0949,
+      "step": 3864
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.49817396635833e-06,
+      "loss": 0.1001,
+      "step": 3865
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.498040679231201e-06,
+      "loss": 0.16,
+      "step": 3866
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.49790739210407e-06,
+      "loss": 0.1245,
+      "step": 3867
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.497774104976942e-06,
+      "loss": 0.1166,
+      "step": 3868
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.497640817849813e-06,
+      "loss": 0.1649,
+      "step": 3869
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.497507530722684e-06,
+      "loss": 0.1263,
+      "step": 3870
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.497374243595555e-06,
+      "loss": 0.1202,
+      "step": 3871
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.497240956468426e-06,
+      "loss": 0.1197,
+      "step": 3872
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.497107669341295e-06,
+      "loss": 0.1657,
+      "step": 3873
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.496974382214168e-06,
+      "loss": 0.1371,
+      "step": 3874
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.496841095087037e-06,
+      "loss": 0.1337,
+      "step": 3875
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.496707807959908e-06,
+      "loss": 0.0966,
+      "step": 3876
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.496574520832779e-06,
+      "loss": 0.1043,
+      "step": 3877
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.49644123370565e-06,
+      "loss": 0.1091,
+      "step": 3878
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.49630794657852e-06,
+      "loss": 0.0881,
+      "step": 3879
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.496174659451392e-06,
+      "loss": 0.0897,
+      "step": 3880
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.496041372324261e-06,
+      "loss": 0.1471,
+      "step": 3881
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.495908085197132e-06,
+      "loss": 0.1086,
+      "step": 3882
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.495774798070003e-06,
+      "loss": 0.1181,
+      "step": 3883
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.495641510942874e-06,
+      "loss": 0.1344,
+      "step": 3884
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.495508223815745e-06,
+      "loss": 0.1229,
+      "step": 3885
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.495374936688614e-06,
+      "loss": 0.1028,
+      "step": 3886
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.495241649561487e-06,
+      "loss": 0.0991,
+      "step": 3887
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.495108362434356e-06,
+      "loss": 0.11,
+      "step": 3888
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.494975075307227e-06,
+      "loss": 0.1502,
+      "step": 3889
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.494841788180098e-06,
+      "loss": 0.114,
+      "step": 3890
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.49470850105297e-06,
+      "loss": 0.1143,
+      "step": 3891
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.49457521392584e-06,
+      "loss": 0.1556,
+      "step": 3892
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.494441926798711e-06,
+      "loss": 0.1236,
+      "step": 3893
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.49430863967158e-06,
+      "loss": 0.0686,
+      "step": 3894
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.494175352544452e-06,
+      "loss": 0.1307,
+      "step": 3895
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.494042065417323e-06,
+      "loss": 0.1188,
+      "step": 3896
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.493908778290194e-06,
+      "loss": 0.0958,
+      "step": 3897
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.493775491163065e-06,
+      "loss": 0.1182,
+      "step": 3898
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.493642204035936e-06,
+      "loss": 0.1383,
+      "step": 3899
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.493508916908805e-06,
+      "loss": 0.1084,
+      "step": 3900
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.493375629781676e-06,
+      "loss": 0.1001,
+      "step": 3901
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.493242342654547e-06,
+      "loss": 0.1586,
+      "step": 3902
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.493109055527418e-06,
+      "loss": 0.1488,
+      "step": 3903
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.492975768400289e-06,
+      "loss": 0.1235,
+      "step": 3904
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.492842481273158e-06,
+      "loss": 0.1381,
+      "step": 3905
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.492709194146031e-06,
+      "loss": 0.073,
+      "step": 3906
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.4925759070189e-06,
+      "loss": 0.1255,
+      "step": 3907
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.492442619891771e-06,
+      "loss": 0.1199,
+      "step": 3908
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.492309332764642e-06,
+      "loss": 0.1454,
+      "step": 3909
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.492176045637513e-06,
+      "loss": 0.1158,
+      "step": 3910
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.492042758510384e-06,
+      "loss": 0.0841,
+      "step": 3911
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.491909471383255e-06,
+      "loss": 0.1202,
+      "step": 3912
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.491776184256124e-06,
+      "loss": 0.1284,
+      "step": 3913
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.491642897128997e-06,
+      "loss": 0.0858,
+      "step": 3914
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.491509610001866e-06,
+      "loss": 0.0972,
+      "step": 3915
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.491376322874737e-06,
+      "loss": 0.0977,
+      "step": 3916
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.491243035747608e-06,
+      "loss": 0.1348,
+      "step": 3917
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.49110974862048e-06,
+      "loss": 0.1377,
+      "step": 3918
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.49097646149335e-06,
+      "loss": 0.1228,
+      "step": 3919
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.49084317436622e-06,
+      "loss": 0.1302,
+      "step": 3920
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.49070988723909e-06,
+      "loss": 0.1415,
+      "step": 3921
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.490576600111962e-06,
+      "loss": 0.0971,
+      "step": 3922
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.490443312984833e-06,
+      "loss": 0.0967,
+      "step": 3923
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.490310025857704e-06,
+      "loss": 0.1502,
+      "step": 3924
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.490176738730575e-06,
+      "loss": 0.1361,
+      "step": 3925
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.490043451603444e-06,
+      "loss": 0.076,
+      "step": 3926
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.489910164476317e-06,
+      "loss": 0.0815,
+      "step": 3927
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.489776877349186e-06,
+      "loss": 0.1283,
+      "step": 3928
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.489643590222057e-06,
+      "loss": 0.0954,
+      "step": 3929
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.489510303094928e-06,
+      "loss": 0.1196,
+      "step": 3930
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.489377015967799e-06,
+      "loss": 0.0953,
+      "step": 3931
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.48924372884067e-06,
+      "loss": 0.1242,
+      "step": 3932
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.489110441713541e-06,
+      "loss": 0.1232,
+      "step": 3933
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.48897715458641e-06,
+      "loss": 0.1006,
+      "step": 3934
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.488843867459281e-06,
+      "loss": 0.1136,
+      "step": 3935
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.488710580332152e-06,
+      "loss": 0.132,
+      "step": 3936
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.488577293205023e-06,
+      "loss": 0.1288,
+      "step": 3937
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.488444006077894e-06,
+      "loss": 0.126,
+      "step": 3938
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.488310718950765e-06,
+      "loss": 0.1448,
+      "step": 3939
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.488177431823635e-06,
+      "loss": 0.1044,
+      "step": 3940
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.488044144696506e-06,
+      "loss": 0.1256,
+      "step": 3941
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.487910857569377e-06,
+      "loss": 0.0859,
+      "step": 3942
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.487777570442248e-06,
+      "loss": 0.1068,
+      "step": 3943
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.487644283315119e-06,
+      "loss": 0.1484,
+      "step": 3944
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.487510996187988e-06,
+      "loss": 0.0754,
+      "step": 3945
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.48737770906086e-06,
+      "loss": 0.1384,
+      "step": 3946
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.48724442193373e-06,
+      "loss": 0.139,
+      "step": 3947
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.4871111348066e-06,
+      "loss": 0.0932,
+      "step": 3948
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.486977847679472e-06,
+      "loss": 0.126,
+      "step": 3949
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.486844560552343e-06,
+      "loss": 0.1111,
+      "step": 3950
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.486711273425214e-06,
+      "loss": 0.1286,
+      "step": 3951
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.486577986298085e-06,
+      "loss": 0.1115,
+      "step": 3952
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.486444699170954e-06,
+      "loss": 0.1102,
+      "step": 3953
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.486311412043827e-06,
+      "loss": 0.1184,
+      "step": 3954
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.486178124916696e-06,
+      "loss": 0.1265,
+      "step": 3955
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.486044837789567e-06,
+      "loss": 0.1201,
+      "step": 3956
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.485911550662438e-06,
+      "loss": 0.0937,
+      "step": 3957
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.485778263535309e-06,
+      "loss": 0.1279,
+      "step": 3958
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.48564497640818e-06,
+      "loss": 0.0993,
+      "step": 3959
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.48551168928105e-06,
+      "loss": 0.0907,
+      "step": 3960
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.48537840215392e-06,
+      "loss": 0.0948,
+      "step": 3961
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.485245115026791e-06,
+      "loss": 0.0808,
+      "step": 3962
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.485111827899662e-06,
+      "loss": 0.1256,
+      "step": 3963
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.484978540772533e-06,
+      "loss": 0.0897,
+      "step": 3964
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.484845253645404e-06,
+      "loss": 0.1309,
+      "step": 3965
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.484711966518274e-06,
+      "loss": 0.1649,
+      "step": 3966
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.484578679391145e-06,
+      "loss": 0.0932,
+      "step": 3967
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.484445392264016e-06,
+      "loss": 0.1455,
+      "step": 3968
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.484312105136887e-06,
+      "loss": 0.0738,
+      "step": 3969
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.484178818009758e-06,
+      "loss": 0.1055,
+      "step": 3970
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.484045530882629e-06,
+      "loss": 0.152,
+      "step": 3971
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.483912243755498e-06,
+      "loss": 0.0963,
+      "step": 3972
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.48377895662837e-06,
+      "loss": 0.1034,
+      "step": 3973
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.48364566950124e-06,
+      "loss": 0.123,
+      "step": 3974
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.483512382374111e-06,
+      "loss": 0.0947,
+      "step": 3975
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.483379095246982e-06,
+      "loss": 0.0938,
+      "step": 3976
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.483245808119853e-06,
+      "loss": 0.0975,
+      "step": 3977
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.483112520992724e-06,
+      "loss": 0.1592,
+      "step": 3978
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.482979233865593e-06,
+      "loss": 0.1298,
+      "step": 3979
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.482845946738464e-06,
+      "loss": 0.1592,
+      "step": 3980
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.482712659611335e-06,
+      "loss": 0.1235,
+      "step": 3981
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.482579372484206e-06,
+      "loss": 0.1288,
+      "step": 3982
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.482446085357077e-06,
+      "loss": 0.115,
+      "step": 3983
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.482312798229948e-06,
+      "loss": 0.1258,
+      "step": 3984
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.482179511102817e-06,
+      "loss": 0.1234,
+      "step": 3985
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.48204622397569e-06,
+      "loss": 0.1239,
+      "step": 3986
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.48191293684856e-06,
+      "loss": 0.0988,
+      "step": 3987
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.48177964972143e-06,
+      "loss": 0.1216,
+      "step": 3988
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.481646362594301e-06,
+      "loss": 0.1237,
+      "step": 3989
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.481513075467172e-06,
+      "loss": 0.1019,
+      "step": 3990
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.481379788340043e-06,
+      "loss": 0.1068,
+      "step": 3991
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.481246501212914e-06,
+      "loss": 0.083,
+      "step": 3992
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.481113214085784e-06,
+      "loss": 0.0922,
+      "step": 3993
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.480979926958655e-06,
+      "loss": 0.1127,
+      "step": 3994
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.480846639831526e-06,
+      "loss": 0.1234,
+      "step": 3995
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.480713352704397e-06,
+      "loss": 0.1189,
+      "step": 3996
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.480580065577268e-06,
+      "loss": 0.1275,
+      "step": 3997
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.480446778450139e-06,
+      "loss": 0.0902,
+      "step": 3998
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.480313491323008e-06,
+      "loss": 0.1044,
+      "step": 3999
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.480180204195879e-06,
+      "loss": 0.1142,
+      "step": 4000
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.48004691706875e-06,
+      "loss": 0.1245,
+      "step": 4001
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.479913629941621e-06,
+      "loss": 0.1327,
+      "step": 4002
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.479780342814492e-06,
+      "loss": 0.1327,
+      "step": 4003
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.479647055687361e-06,
+      "loss": 0.0986,
+      "step": 4004
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.479513768560234e-06,
+      "loss": 0.1076,
+      "step": 4005
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.479380481433103e-06,
+      "loss": 0.0715,
+      "step": 4006
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.479247194305974e-06,
+      "loss": 0.1333,
+      "step": 4007
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.479113907178845e-06,
+      "loss": 0.1127,
+      "step": 4008
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.478980620051716e-06,
+      "loss": 0.0956,
+      "step": 4009
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.478847332924587e-06,
+      "loss": 0.1055,
+      "step": 4010
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.478714045797458e-06,
+      "loss": 0.0943,
+      "step": 4011
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.478580758670328e-06,
+      "loss": 0.1042,
+      "step": 4012
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.4784474715432e-06,
+      "loss": 0.1295,
+      "step": 4013
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.47831418441607e-06,
+      "loss": 0.1358,
+      "step": 4014
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.47818089728894e-06,
+      "loss": 0.0958,
+      "step": 4015
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.478047610161812e-06,
+      "loss": 0.1225,
+      "step": 4016
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.477914323034683e-06,
+      "loss": 0.1111,
+      "step": 4017
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.477781035907553e-06,
+      "loss": 0.0982,
+      "step": 4018
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.477647748780423e-06,
+      "loss": 0.1692,
+      "step": 4019
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.477514461653294e-06,
+      "loss": 0.0923,
+      "step": 4020
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.477381174526165e-06,
+      "loss": 0.1057,
+      "step": 4021
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.477247887399036e-06,
+      "loss": 0.0944,
+      "step": 4022
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.477114600271907e-06,
+      "loss": 0.1262,
+      "step": 4023
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.476981313144778e-06,
+      "loss": 0.1126,
+      "step": 4024
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.476848026017647e-06,
+      "loss": 0.1467,
+      "step": 4025
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.47671473889052e-06,
+      "loss": 0.0956,
+      "step": 4026
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.476581451763389e-06,
+      "loss": 0.1078,
+      "step": 4027
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.47644816463626e-06,
+      "loss": 0.1045,
+      "step": 4028
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.476314877509131e-06,
+      "loss": 0.1055,
+      "step": 4029
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.476181590382002e-06,
+      "loss": 0.1274,
+      "step": 4030
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.476048303254873e-06,
+      "loss": 0.1113,
+      "step": 4031
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.475915016127744e-06,
+      "loss": 0.1554,
+      "step": 4032
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.475781729000613e-06,
+      "loss": 0.1212,
+      "step": 4033
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.475648441873484e-06,
+      "loss": 0.136,
+      "step": 4034
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.475515154746355e-06,
+      "loss": 0.1006,
+      "step": 4035
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.475381867619226e-06,
+      "loss": 0.1074,
+      "step": 4036
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.475248580492097e-06,
+      "loss": 0.1236,
+      "step": 4037
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.475115293364967e-06,
+      "loss": 0.1274,
+      "step": 4038
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.474982006237838e-06,
+      "loss": 0.1199,
+      "step": 4039
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.474848719110709e-06,
+      "loss": 0.0884,
+      "step": 4040
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.47471543198358e-06,
+      "loss": 0.1144,
+      "step": 4041
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.47458214485645e-06,
+      "loss": 0.1489,
+      "step": 4042
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.474448857729322e-06,
+      "loss": 0.1257,
+      "step": 4043
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.474315570602191e-06,
+      "loss": 0.0759,
+      "step": 4044
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.474182283475064e-06,
+      "loss": 0.1337,
+      "step": 4045
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.474048996347933e-06,
+      "loss": 0.1055,
+      "step": 4046
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.473915709220804e-06,
+      "loss": 0.1183,
+      "step": 4047
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.473782422093675e-06,
+      "loss": 0.1399,
+      "step": 4048
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.473649134966546e-06,
+      "loss": 0.1115,
+      "step": 4049
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.473515847839417e-06,
+      "loss": 0.118,
+      "step": 4050
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.473382560712288e-06,
+      "loss": 0.0985,
+      "step": 4051
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.473249273585157e-06,
+      "loss": 0.0992,
+      "step": 4052
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.47311598645803e-06,
+      "loss": 0.1543,
+      "step": 4053
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.472982699330899e-06,
+      "loss": 0.1055,
+      "step": 4054
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.47284941220377e-06,
+      "loss": 0.189,
+      "step": 4055
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.472716125076641e-06,
+      "loss": 0.1125,
+      "step": 4056
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.47258283794951e-06,
+      "loss": 0.1192,
+      "step": 4057
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.472449550822383e-06,
+      "loss": 0.1321,
+      "step": 4058
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.472316263695252e-06,
+      "loss": 0.1164,
+      "step": 4059
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.472182976568123e-06,
+      "loss": 0.1164,
+      "step": 4060
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.472049689440994e-06,
+      "loss": 0.1186,
+      "step": 4061
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.471916402313865e-06,
+      "loss": 0.1018,
+      "step": 4062
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.471783115186736e-06,
+      "loss": 0.0989,
+      "step": 4063
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.471649828059607e-06,
+      "loss": 0.1116,
+      "step": 4064
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.471516540932477e-06,
+      "loss": 0.0886,
+      "step": 4065
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.471383253805348e-06,
+      "loss": 0.1137,
+      "step": 4066
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.471249966678219e-06,
+      "loss": 0.0922,
+      "step": 4067
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.47111667955109e-06,
+      "loss": 0.1129,
+      "step": 4068
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.47098339242396e-06,
+      "loss": 0.0979,
+      "step": 4069
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.470850105296832e-06,
+      "loss": 0.1157,
+      "step": 4070
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.470716818169701e-06,
+      "loss": 0.1398,
+      "step": 4071
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.470583531042574e-06,
+      "loss": 0.1074,
+      "step": 4072
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.470450243915443e-06,
+      "loss": 0.1416,
+      "step": 4073
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.470316956788314e-06,
+      "loss": 0.1363,
+      "step": 4074
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.470183669661185e-06,
+      "loss": 0.0978,
+      "step": 4075
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.470050382534056e-06,
+      "loss": 0.1417,
+      "step": 4076
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.469917095406927e-06,
+      "loss": 0.1465,
+      "step": 4077
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.469783808279796e-06,
+      "loss": 0.1156,
+      "step": 4078
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.469650521152667e-06,
+      "loss": 0.1069,
+      "step": 4079
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.469517234025538e-06,
+      "loss": 0.1533,
+      "step": 4080
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.46938394689841e-06,
+      "loss": 0.1097,
+      "step": 4081
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.46925065977128e-06,
+      "loss": 0.1207,
+      "step": 4082
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.469117372644151e-06,
+      "loss": 0.0978,
+      "step": 4083
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.46898408551702e-06,
+      "loss": 0.0977,
+      "step": 4084
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.468850798389893e-06,
+      "loss": 0.101,
+      "step": 4085
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.468717511262763e-06,
+      "loss": 0.0865,
+      "step": 4086
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.468584224135633e-06,
+      "loss": 0.1076,
+      "step": 4087
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.468450937008504e-06,
+      "loss": 0.0889,
+      "step": 4088
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.468317649881375e-06,
+      "loss": 0.0926,
+      "step": 4089
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.468184362754246e-06,
+      "loss": 0.1196,
+      "step": 4090
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.468051075627117e-06,
+      "loss": 0.1109,
+      "step": 4091
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.467917788499987e-06,
+      "loss": 0.1044,
+      "step": 4092
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.46778450137286e-06,
+      "loss": 0.1459,
+      "step": 4093
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.467651214245729e-06,
+      "loss": 0.1211,
+      "step": 4094
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.4675179271186e-06,
+      "loss": 0.1391,
+      "step": 4095
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.46738463999147e-06,
+      "loss": 0.1149,
+      "step": 4096
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.46725135286434e-06,
+      "loss": 0.1477,
+      "step": 4097
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.467118065737213e-06,
+      "loss": 0.1187,
+      "step": 4098
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.466984778610082e-06,
+      "loss": 0.1103,
+      "step": 4099
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.466851491482953e-06,
+      "loss": 0.0893,
+      "step": 4100
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.466718204355824e-06,
+      "loss": 0.0978,
+      "step": 4101
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.466584917228695e-06,
+      "loss": 0.0934,
+      "step": 4102
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.466451630101566e-06,
+      "loss": 0.1154,
+      "step": 4103
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.466318342974437e-06,
+      "loss": 0.0836,
+      "step": 4104
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.466185055847306e-06,
+      "loss": 0.1781,
+      "step": 4105
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.466051768720177e-06,
+      "loss": 0.1261,
+      "step": 4106
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.465918481593048e-06,
+      "loss": 0.1187,
+      "step": 4107
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.46578519446592e-06,
+      "loss": 0.1338,
+      "step": 4108
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.46565190733879e-06,
+      "loss": 0.1082,
+      "step": 4109
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.465518620211661e-06,
+      "loss": 0.1104,
+      "step": 4110
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.46538533308453e-06,
+      "loss": 0.0856,
+      "step": 4111
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.465252045957403e-06,
+      "loss": 0.1384,
+      "step": 4112
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.465118758830273e-06,
+      "loss": 0.1213,
+      "step": 4113
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.464985471703144e-06,
+      "loss": 0.0767,
+      "step": 4114
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.464852184576015e-06,
+      "loss": 0.1305,
+      "step": 4115
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.464718897448884e-06,
+      "loss": 0.0808,
+      "step": 4116
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.464585610321757e-06,
+      "loss": 0.1143,
+      "step": 4117
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.464452323194626e-06,
+      "loss": 0.1018,
+      "step": 4118
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.464319036067497e-06,
+      "loss": 0.1301,
+      "step": 4119
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.464185748940368e-06,
+      "loss": 0.1223,
+      "step": 4120
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.464052461813239e-06,
+      "loss": 0.1341,
+      "step": 4121
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.46391917468611e-06,
+      "loss": 0.0893,
+      "step": 4122
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.46378588755898e-06,
+      "loss": 0.1241,
+      "step": 4123
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.46365260043185e-06,
+      "loss": 0.091,
+      "step": 4124
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.463519313304723e-06,
+      "loss": 0.1201,
+      "step": 4125
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.463386026177592e-06,
+      "loss": 0.1326,
+      "step": 4126
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.463252739050463e-06,
+      "loss": 0.1541,
+      "step": 4127
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.463119451923334e-06,
+      "loss": 0.0904,
+      "step": 4128
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.462986164796205e-06,
+      "loss": 0.1223,
+      "step": 4129
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.462852877669076e-06,
+      "loss": 0.1129,
+      "step": 4130
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.462719590541947e-06,
+      "loss": 0.1444,
+      "step": 4131
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.462586303414816e-06,
+      "loss": 0.1655,
+      "step": 4132
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.462453016287687e-06,
+      "loss": 0.1469,
+      "step": 4133
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.462319729160558e-06,
+      "loss": 0.1101,
+      "step": 4134
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.46218644203343e-06,
+      "loss": 0.1195,
+      "step": 4135
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.4620531549063e-06,
+      "loss": 0.1195,
+      "step": 4136
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.46191986777917e-06,
+      "loss": 0.1255,
+      "step": 4137
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.46178658065204e-06,
+      "loss": 0.1122,
+      "step": 4138
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.461653293524912e-06,
+      "loss": 0.166,
+      "step": 4139
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.461520006397783e-06,
+      "loss": 0.1066,
+      "step": 4140
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.461386719270654e-06,
+      "loss": 0.1379,
+      "step": 4141
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.461253432143525e-06,
+      "loss": 0.0993,
+      "step": 4142
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.461120145016394e-06,
+      "loss": 0.0955,
+      "step": 4143
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.460986857889267e-06,
+      "loss": 0.1377,
+      "step": 4144
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.460853570762136e-06,
+      "loss": 0.1834,
+      "step": 4145
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.460720283635007e-06,
+      "loss": 0.0767,
+      "step": 4146
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.460586996507878e-06,
+      "loss": 0.1103,
+      "step": 4147
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.460453709380749e-06,
+      "loss": 0.0999,
+      "step": 4148
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.46032042225362e-06,
+      "loss": 0.1098,
+      "step": 4149
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.460187135126491e-06,
+      "loss": 0.0991,
+      "step": 4150
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.46005384799936e-06,
+      "loss": 0.109,
+      "step": 4151
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.459920560872233e-06,
+      "loss": 0.1153,
+      "step": 4152
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.459787273745102e-06,
+      "loss": 0.0885,
+      "step": 4153
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.459653986617973e-06,
+      "loss": 0.1117,
+      "step": 4154
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.459520699490844e-06,
+      "loss": 0.1406,
+      "step": 4155
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.459387412363713e-06,
+      "loss": 0.1195,
+      "step": 4156
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.459254125236586e-06,
+      "loss": 0.0948,
+      "step": 4157
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.459120838109455e-06,
+      "loss": 0.1328,
+      "step": 4158
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.458987550982326e-06,
+      "loss": 0.1077,
+      "step": 4159
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.458854263855197e-06,
+      "loss": 0.1226,
+      "step": 4160
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.458720976728068e-06,
+      "loss": 0.1135,
+      "step": 4161
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.45858768960094e-06,
+      "loss": 0.0735,
+      "step": 4162
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.45845440247381e-06,
+      "loss": 0.1361,
+      "step": 4163
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.45832111534668e-06,
+      "loss": 0.1133,
+      "step": 4164
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.458187828219552e-06,
+      "loss": 0.1324,
+      "step": 4165
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.458054541092422e-06,
+      "loss": 0.1041,
+      "step": 4166
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.457921253965293e-06,
+      "loss": 0.1292,
+      "step": 4167
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.457787966838164e-06,
+      "loss": 0.1253,
+      "step": 4168
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.457654679711035e-06,
+      "loss": 0.1268,
+      "step": 4169
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.457521392583906e-06,
+      "loss": 0.0935,
+      "step": 4170
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.457388105456777e-06,
+      "loss": 0.1571,
+      "step": 4171
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.457254818329646e-06,
+      "loss": 0.1512,
+      "step": 4172
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.457121531202517e-06,
+      "loss": 0.1244,
+      "step": 4173
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.456988244075388e-06,
+      "loss": 0.1435,
+      "step": 4174
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.456854956948259e-06,
+      "loss": 0.1674,
+      "step": 4175
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.45672166982113e-06,
+      "loss": 0.1232,
+      "step": 4176
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.456588382694e-06,
+      "loss": 0.0968,
+      "step": 4177
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.45645509556687e-06,
+      "loss": 0.1291,
+      "step": 4178
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.456321808439741e-06,
+      "loss": 0.1151,
+      "step": 4179
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.456188521312612e-06,
+      "loss": 0.0859,
+      "step": 4180
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.456055234185483e-06,
+      "loss": 0.1277,
+      "step": 4181
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.455921947058354e-06,
+      "loss": 0.1215,
+      "step": 4182
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.455788659931224e-06,
+      "loss": 0.1547,
+      "step": 4183
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.455655372804096e-06,
+      "loss": 0.0955,
+      "step": 4184
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.455522085676966e-06,
+      "loss": 0.1086,
+      "step": 4185
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.455388798549837e-06,
+      "loss": 0.1701,
+      "step": 4186
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.455255511422708e-06,
+      "loss": 0.1039,
+      "step": 4187
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.455122224295579e-06,
+      "loss": 0.072,
+      "step": 4188
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.45498893716845e-06,
+      "loss": 0.1265,
+      "step": 4189
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.45485565004132e-06,
+      "loss": 0.1107,
+      "step": 4190
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.45472236291419e-06,
+      "loss": 0.1276,
+      "step": 4191
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.454589075787063e-06,
+      "loss": 0.1378,
+      "step": 4192
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.454455788659932e-06,
+      "loss": 0.0978,
+      "step": 4193
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.454322501532803e-06,
+      "loss": 0.1097,
+      "step": 4194
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.454189214405674e-06,
+      "loss": 0.1082,
+      "step": 4195
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.454055927278543e-06,
+      "loss": 0.1525,
+      "step": 4196
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.453922640151416e-06,
+      "loss": 0.1185,
+      "step": 4197
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.453789353024285e-06,
+      "loss": 0.1032,
+      "step": 4198
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.453656065897156e-06,
+      "loss": 0.0902,
+      "step": 4199
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.453522778770027e-06,
+      "loss": 0.1146,
+      "step": 4200
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.453389491642898e-06,
+      "loss": 0.1581,
+      "step": 4201
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.453256204515769e-06,
+      "loss": 0.1641,
+      "step": 4202
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.45312291738864e-06,
+      "loss": 0.1324,
+      "step": 4203
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.45298963026151e-06,
+      "loss": 0.1616,
+      "step": 4204
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.45285634313438e-06,
+      "loss": 0.1098,
+      "step": 4205
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.452723056007251e-06,
+      "loss": 0.1397,
+      "step": 4206
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.452589768880122e-06,
+      "loss": 0.1253,
+      "step": 4207
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.452456481752993e-06,
+      "loss": 0.1096,
+      "step": 4208
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.452323194625864e-06,
+      "loss": 0.1092,
+      "step": 4209
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.452189907498734e-06,
+      "loss": 0.121,
+      "step": 4210
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.452056620371606e-06,
+      "loss": 0.1187,
+      "step": 4211
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.451923333244476e-06,
+      "loss": 0.1351,
+      "step": 4212
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.451790046117347e-06,
+      "loss": 0.1287,
+      "step": 4213
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.451656758990218e-06,
+      "loss": 0.1163,
+      "step": 4214
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.451523471863087e-06,
+      "loss": 0.118,
+      "step": 4215
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.45139018473596e-06,
+      "loss": 0.1196,
+      "step": 4216
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.451256897608829e-06,
+      "loss": 0.1353,
+      "step": 4217
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.4511236104817e-06,
+      "loss": 0.1337,
+      "step": 4218
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.450990323354571e-06,
+      "loss": 0.0921,
+      "step": 4219
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.450857036227442e-06,
+      "loss": 0.1488,
+      "step": 4220
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.450723749100313e-06,
+      "loss": 0.0803,
+      "step": 4221
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.450590461973184e-06,
+      "loss": 0.108,
+      "step": 4222
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.450457174846053e-06,
+      "loss": 0.1361,
+      "step": 4223
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.450323887718926e-06,
+      "loss": 0.1074,
+      "step": 4224
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.450190600591795e-06,
+      "loss": 0.0801,
+      "step": 4225
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.450057313464666e-06,
+      "loss": 0.1216,
+      "step": 4226
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.449924026337537e-06,
+      "loss": 0.1509,
+      "step": 4227
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.449790739210408e-06,
+      "loss": 0.1046,
+      "step": 4228
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.44965745208328e-06,
+      "loss": 0.0977,
+      "step": 4229
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.44952416495615e-06,
+      "loss": 0.0834,
+      "step": 4230
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.44939087782902e-06,
+      "loss": 0.1232,
+      "step": 4231
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.44925759070189e-06,
+      "loss": 0.1062,
+      "step": 4232
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.449124303574761e-06,
+      "loss": 0.0821,
+      "step": 4233
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.448991016447632e-06,
+      "loss": 0.0978,
+      "step": 4234
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.448857729320503e-06,
+      "loss": 0.1243,
+      "step": 4235
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.448724442193373e-06,
+      "loss": 0.178,
+      "step": 4236
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.448591155066244e-06,
+      "loss": 0.127,
+      "step": 4237
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.448457867939115e-06,
+      "loss": 0.1588,
+      "step": 4238
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.448324580811986e-06,
+      "loss": 0.0953,
+      "step": 4239
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.448191293684857e-06,
+      "loss": 0.1072,
+      "step": 4240
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.448058006557728e-06,
+      "loss": 0.0931,
+      "step": 4241
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.447924719430597e-06,
+      "loss": 0.1183,
+      "step": 4242
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.44779143230347e-06,
+      "loss": 0.106,
+      "step": 4243
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.447658145176339e-06,
+      "loss": 0.1472,
+      "step": 4244
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.44752485804921e-06,
+      "loss": 0.1154,
+      "step": 4245
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.447391570922081e-06,
+      "loss": 0.1137,
+      "step": 4246
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.447258283794952e-06,
+      "loss": 0.1079,
+      "step": 4247
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.447124996667823e-06,
+      "loss": 0.0986,
+      "step": 4248
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.446991709540694e-06,
+      "loss": 0.1,
+      "step": 4249
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.446858422413563e-06,
+      "loss": 0.1219,
+      "step": 4250
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.446725135286436e-06,
+      "loss": 0.1167,
+      "step": 4251
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.446591848159305e-06,
+      "loss": 0.118,
+      "step": 4252
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.446458561032176e-06,
+      "loss": 0.0836,
+      "step": 4253
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.446325273905047e-06,
+      "loss": 0.117,
+      "step": 4254
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.446191986777917e-06,
+      "loss": 0.1029,
+      "step": 4255
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.44605869965079e-06,
+      "loss": 0.1129,
+      "step": 4256
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.445925412523659e-06,
+      "loss": 0.1329,
+      "step": 4257
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.44579212539653e-06,
+      "loss": 0.1348,
+      "step": 4258
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.4456588382694e-06,
+      "loss": 0.1273,
+      "step": 4259
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.445525551142272e-06,
+      "loss": 0.0751,
+      "step": 4260
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.445392264015143e-06,
+      "loss": 0.0969,
+      "step": 4261
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.445258976888014e-06,
+      "loss": 0.1072,
+      "step": 4262
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.445125689760883e-06,
+      "loss": 0.0864,
+      "step": 4263
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.444992402633756e-06,
+      "loss": 0.1075,
+      "step": 4264
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.444859115506625e-06,
+      "loss": 0.0921,
+      "step": 4265
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.444725828379496e-06,
+      "loss": 0.1551,
+      "step": 4266
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.444592541252367e-06,
+      "loss": 0.1366,
+      "step": 4267
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.444459254125238e-06,
+      "loss": 0.1575,
+      "step": 4268
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.444325966998109e-06,
+      "loss": 0.0841,
+      "step": 4269
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.44419267987098e-06,
+      "loss": 0.1502,
+      "step": 4270
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.444059392743849e-06,
+      "loss": 0.1057,
+      "step": 4271
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.44392610561672e-06,
+      "loss": 0.1591,
+      "step": 4272
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.443792818489591e-06,
+      "loss": 0.1015,
+      "step": 4273
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.443659531362462e-06,
+      "loss": 0.1385,
+      "step": 4274
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.443526244235333e-06,
+      "loss": 0.0894,
+      "step": 4275
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.443392957108202e-06,
+      "loss": 0.0739,
+      "step": 4276
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.443259669981073e-06,
+      "loss": 0.1337,
+      "step": 4277
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.443126382853944e-06,
+      "loss": 0.113,
+      "step": 4278
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.442993095726815e-06,
+      "loss": 0.113,
+      "step": 4279
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.442859808599686e-06,
+      "loss": 0.1316,
+      "step": 4280
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.442726521472557e-06,
+      "loss": 0.1108,
+      "step": 4281
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.442593234345427e-06,
+      "loss": 0.1263,
+      "step": 4282
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.4424599472183e-06,
+      "loss": 0.0876,
+      "step": 4283
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.442326660091169e-06,
+      "loss": 0.1523,
+      "step": 4284
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.44219337296404e-06,
+      "loss": 0.1486,
+      "step": 4285
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.44206008583691e-06,
+      "loss": 0.1575,
+      "step": 4286
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.441926798709782e-06,
+      "loss": 0.1328,
+      "step": 4287
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.441793511582653e-06,
+      "loss": 0.1623,
+      "step": 4288
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.441660224455524e-06,
+      "loss": 0.0796,
+      "step": 4289
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.441526937328393e-06,
+      "loss": 0.1175,
+      "step": 4290
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.441393650201266e-06,
+      "loss": 0.1222,
+      "step": 4291
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.441260363074135e-06,
+      "loss": 0.1032,
+      "step": 4292
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.441127075947006e-06,
+      "loss": 0.1418,
+      "step": 4293
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.440993788819877e-06,
+      "loss": 0.0863,
+      "step": 4294
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.440860501692746e-06,
+      "loss": 0.113,
+      "step": 4295
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.440727214565619e-06,
+      "loss": 0.0878,
+      "step": 4296
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.440593927438488e-06,
+      "loss": 0.1394,
+      "step": 4297
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.44046064031136e-06,
+      "loss": 0.0942,
+      "step": 4298
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.44032735318423e-06,
+      "loss": 0.1098,
+      "step": 4299
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.440194066057101e-06,
+      "loss": 0.0977,
+      "step": 4300
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.440060778929972e-06,
+      "loss": 0.1054,
+      "step": 4301
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.439927491802843e-06,
+      "loss": 0.1179,
+      "step": 4302
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.439794204675712e-06,
+      "loss": 0.1027,
+      "step": 4303
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.439660917548583e-06,
+      "loss": 0.1146,
+      "step": 4304
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.439527630421454e-06,
+      "loss": 0.1403,
+      "step": 4305
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.439394343294325e-06,
+      "loss": 0.1538,
+      "step": 4306
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.439261056167196e-06,
+      "loss": 0.1972,
+      "step": 4307
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.439127769040067e-06,
+      "loss": 0.0647,
+      "step": 4308
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.438994481912937e-06,
+      "loss": 0.1266,
+      "step": 4309
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.43886119478581e-06,
+      "loss": 0.1224,
+      "step": 4310
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.438727907658679e-06,
+      "loss": 0.0982,
+      "step": 4311
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.43859462053155e-06,
+      "loss": 0.106,
+      "step": 4312
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.43846133340442e-06,
+      "loss": 0.1531,
+      "step": 4313
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.43832804627729e-06,
+      "loss": 0.0908,
+      "step": 4314
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.438194759150163e-06,
+      "loss": 0.1143,
+      "step": 4315
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.438061472023032e-06,
+      "loss": 0.0952,
+      "step": 4316
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.437928184895903e-06,
+      "loss": 0.1094,
+      "step": 4317
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.437794897768774e-06,
+      "loss": 0.0992,
+      "step": 4318
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.437661610641645e-06,
+      "loss": 0.0938,
+      "step": 4319
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.437528323514516e-06,
+      "loss": 0.1349,
+      "step": 4320
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.437395036387387e-06,
+      "loss": 0.1119,
+      "step": 4321
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.437261749260256e-06,
+      "loss": 0.1015,
+      "step": 4322
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.437128462133129e-06,
+      "loss": 0.1471,
+      "step": 4323
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.436995175005998e-06,
+      "loss": 0.1361,
+      "step": 4324
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.43686188787887e-06,
+      "loss": 0.1159,
+      "step": 4325
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.43672860075174e-06,
+      "loss": 0.1151,
+      "step": 4326
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.436595313624611e-06,
+      "loss": 0.1141,
+      "step": 4327
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.436462026497482e-06,
+      "loss": 0.1277,
+      "step": 4328
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.436328739370353e-06,
+      "loss": 0.1175,
+      "step": 4329
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.436195452243223e-06,
+      "loss": 0.1298,
+      "step": 4330
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.436062165116095e-06,
+      "loss": 0.0632,
+      "step": 4331
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.435928877988965e-06,
+      "loss": 0.0846,
+      "step": 4332
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.435795590861836e-06,
+      "loss": 0.0867,
+      "step": 4333
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.435662303734707e-06,
+      "loss": 0.0732,
+      "step": 4334
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.435529016607576e-06,
+      "loss": 0.1349,
+      "step": 4335
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.435395729480448e-06,
+      "loss": 0.1215,
+      "step": 4336
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.435262442353318e-06,
+      "loss": 0.1189,
+      "step": 4337
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.435129155226189e-06,
+      "loss": 0.119,
+      "step": 4338
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.43499586809906e-06,
+      "loss": 0.143,
+      "step": 4339
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.43486258097193e-06,
+      "loss": 0.1077,
+      "step": 4340
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.434729293844802e-06,
+      "loss": 0.134,
+      "step": 4341
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.434596006717673e-06,
+      "loss": 0.1158,
+      "step": 4342
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.434462719590542e-06,
+      "loss": 0.1206,
+      "step": 4343
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.434329432463413e-06,
+      "loss": 0.128,
+      "step": 4344
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.434196145336284e-06,
+      "loss": 0.1341,
+      "step": 4345
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.434062858209155e-06,
+      "loss": 0.1754,
+      "step": 4346
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.433929571082026e-06,
+      "loss": 0.1366,
+      "step": 4347
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.433796283954897e-06,
+      "loss": 0.0941,
+      "step": 4348
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.433662996827766e-06,
+      "loss": 0.1472,
+      "step": 4349
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.433529709700639e-06,
+      "loss": 0.1373,
+      "step": 4350
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.433396422573508e-06,
+      "loss": 0.1071,
+      "step": 4351
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.43326313544638e-06,
+      "loss": 0.1504,
+      "step": 4352
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.43312984831925e-06,
+      "loss": 0.0992,
+      "step": 4353
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.43299656119212e-06,
+      "loss": 0.1399,
+      "step": 4354
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.432863274064992e-06,
+      "loss": 0.1207,
+      "step": 4355
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.432729986937862e-06,
+      "loss": 0.1423,
+      "step": 4356
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.432596699810733e-06,
+      "loss": 0.1441,
+      "step": 4357
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.432463412683604e-06,
+      "loss": 0.098,
+      "step": 4358
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.432330125556475e-06,
+      "loss": 0.1005,
+      "step": 4359
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.432196838429346e-06,
+      "loss": 0.0937,
+      "step": 4360
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.432063551302217e-06,
+      "loss": 0.0986,
+      "step": 4361
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.431930264175086e-06,
+      "loss": 0.1419,
+      "step": 4362
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.431796977047959e-06,
+      "loss": 0.0959,
+      "step": 4363
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.431663689920828e-06,
+      "loss": 0.1147,
+      "step": 4364
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.431530402793699e-06,
+      "loss": 0.0953,
+      "step": 4365
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.43139711566657e-06,
+      "loss": 0.1144,
+      "step": 4366
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.431263828539441e-06,
+      "loss": 0.0813,
+      "step": 4367
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.431130541412312e-06,
+      "loss": 0.1358,
+      "step": 4368
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.430997254285183e-06,
+      "loss": 0.1442,
+      "step": 4369
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.430863967158052e-06,
+      "loss": 0.0974,
+      "step": 4370
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.430730680030923e-06,
+      "loss": 0.1337,
+      "step": 4371
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.430597392903794e-06,
+      "loss": 0.1027,
+      "step": 4372
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.430464105776665e-06,
+      "loss": 0.0903,
+      "step": 4373
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.430330818649536e-06,
+      "loss": 0.1128,
+      "step": 4374
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.430197531522405e-06,
+      "loss": 0.1056,
+      "step": 4375
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.430064244395276e-06,
+      "loss": 0.0838,
+      "step": 4376
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.429930957268147e-06,
+      "loss": 0.1147,
+      "step": 4377
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.429797670141018e-06,
+      "loss": 0.101,
+      "step": 4378
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.42966438301389e-06,
+      "loss": 0.0979,
+      "step": 4379
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.42953109588676e-06,
+      "loss": 0.1144,
+      "step": 4380
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.42939780875963e-06,
+      "loss": 0.0977,
+      "step": 4381
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.429264521632502e-06,
+      "loss": 0.1192,
+      "step": 4382
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.429131234505372e-06,
+      "loss": 0.1589,
+      "step": 4383
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.428997947378243e-06,
+      "loss": 0.1145,
+      "step": 4384
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.428864660251114e-06,
+      "loss": 0.0983,
+      "step": 4385
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.428731373123985e-06,
+      "loss": 0.136,
+      "step": 4386
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.428598085996856e-06,
+      "loss": 0.1347,
+      "step": 4387
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.428464798869727e-06,
+      "loss": 0.1031,
+      "step": 4388
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.428331511742596e-06,
+      "loss": 0.1252,
+      "step": 4389
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.428198224615467e-06,
+      "loss": 0.1653,
+      "step": 4390
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.428064937488338e-06,
+      "loss": 0.0891,
+      "step": 4391
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.427931650361209e-06,
+      "loss": 0.108,
+      "step": 4392
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.42779836323408e-06,
+      "loss": 0.1543,
+      "step": 4393
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.42766507610695e-06,
+      "loss": 0.1608,
+      "step": 4394
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.427531788979822e-06,
+      "loss": 0.1021,
+      "step": 4395
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.427398501852691e-06,
+      "loss": 0.1389,
+      "step": 4396
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.427265214725562e-06,
+      "loss": 0.1132,
+      "step": 4397
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.427131927598433e-06,
+      "loss": 0.1036,
+      "step": 4398
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.426998640471304e-06,
+      "loss": 0.1409,
+      "step": 4399
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.426865353344175e-06,
+      "loss": 0.1242,
+      "step": 4400
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.426732066217046e-06,
+      "loss": 0.1142,
+      "step": 4401
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.426598779089916e-06,
+      "loss": 0.1032,
+      "step": 4402
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.426465491962788e-06,
+      "loss": 0.1104,
+      "step": 4403
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.426332204835658e-06,
+      "loss": 0.0933,
+      "step": 4404
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.426198917708528e-06,
+      "loss": 0.1148,
+      "step": 4405
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.4260656305814e-06,
+      "loss": 0.1198,
+      "step": 4406
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.42593234345427e-06,
+      "loss": 0.0892,
+      "step": 4407
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.425799056327141e-06,
+      "loss": 0.1091,
+      "step": 4408
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.425665769200012e-06,
+      "loss": 0.1274,
+      "step": 4409
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.425532482072882e-06,
+      "loss": 0.0851,
+      "step": 4410
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.425399194945753e-06,
+      "loss": 0.115,
+      "step": 4411
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.425265907818624e-06,
+      "loss": 0.1407,
+      "step": 4412
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.425132620691495e-06,
+      "loss": 0.1073,
+      "step": 4413
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.424999333564366e-06,
+      "loss": 0.1161,
+      "step": 4414
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.424866046437235e-06,
+      "loss": 0.1078,
+      "step": 4415
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.424732759310106e-06,
+      "loss": 0.0933,
+      "step": 4416
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.424599472182977e-06,
+      "loss": 0.0875,
+      "step": 4417
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.424466185055848e-06,
+      "loss": 0.122,
+      "step": 4418
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.424332897928719e-06,
+      "loss": 0.1602,
+      "step": 4419
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.42419961080159e-06,
+      "loss": 0.1218,
+      "step": 4420
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.42406632367446e-06,
+      "loss": 0.0731,
+      "step": 4421
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.423933036547332e-06,
+      "loss": 0.0957,
+      "step": 4422
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.423799749420201e-06,
+      "loss": 0.0999,
+      "step": 4423
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.423666462293072e-06,
+      "loss": 0.1456,
+      "step": 4424
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.423533175165943e-06,
+      "loss": 0.1463,
+      "step": 4425
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.423399888038814e-06,
+      "loss": 0.2008,
+      "step": 4426
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.423266600911685e-06,
+      "loss": 0.1396,
+      "step": 4427
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.423133313784556e-06,
+      "loss": 0.1099,
+      "step": 4428
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.423000026657426e-06,
+      "loss": 0.1371,
+      "step": 4429
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.422866739530297e-06,
+      "loss": 0.1081,
+      "step": 4430
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.422733452403168e-06,
+      "loss": 0.1216,
+      "step": 4431
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.422600165276039e-06,
+      "loss": 0.1383,
+      "step": 4432
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.42246687814891e-06,
+      "loss": 0.1221,
+      "step": 4433
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.422333591021779e-06,
+      "loss": 0.0956,
+      "step": 4434
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.422200303894652e-06,
+      "loss": 0.1003,
+      "step": 4435
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.422067016767521e-06,
+      "loss": 0.1451,
+      "step": 4436
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.421933729640392e-06,
+      "loss": 0.1778,
+      "step": 4437
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.421800442513263e-06,
+      "loss": 0.1145,
+      "step": 4438
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.421667155386134e-06,
+      "loss": 0.1056,
+      "step": 4439
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.421533868259005e-06,
+      "loss": 0.1238,
+      "step": 4440
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.421400581131876e-06,
+      "loss": 0.0981,
+      "step": 4441
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.421267294004745e-06,
+      "loss": 0.0971,
+      "step": 4442
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.421134006877616e-06,
+      "loss": 0.1038,
+      "step": 4443
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.421000719750487e-06,
+      "loss": 0.1408,
+      "step": 4444
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.420867432623358e-06,
+      "loss": 0.1327,
+      "step": 4445
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.420734145496229e-06,
+      "loss": 0.1482,
+      "step": 4446
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.4206008583691e-06,
+      "loss": 0.1171,
+      "step": 4447
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.42046757124197e-06,
+      "loss": 0.0753,
+      "step": 4448
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.42033428411484e-06,
+      "loss": 0.0879,
+      "step": 4449
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.420200996987711e-06,
+      "loss": 0.1145,
+      "step": 4450
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.420067709860582e-06,
+      "loss": 0.1698,
+      "step": 4451
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.419934422733453e-06,
+      "loss": 0.1001,
+      "step": 4452
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.419801135606323e-06,
+      "loss": 0.1106,
+      "step": 4453
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.419667848479195e-06,
+      "loss": 0.072,
+      "step": 4454
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.419534561352065e-06,
+      "loss": 0.1529,
+      "step": 4455
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.419401274224936e-06,
+      "loss": 0.0975,
+      "step": 4456
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.419267987097807e-06,
+      "loss": 0.0978,
+      "step": 4457
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.419134699970678e-06,
+      "loss": 0.1262,
+      "step": 4458
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.419001412843549e-06,
+      "loss": 0.087,
+      "step": 4459
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.41886812571642e-06,
+      "loss": 0.1225,
+      "step": 4460
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.418734838589289e-06,
+      "loss": 0.1306,
+      "step": 4461
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.418601551462162e-06,
+      "loss": 0.1134,
+      "step": 4462
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.418468264335031e-06,
+      "loss": 0.113,
+      "step": 4463
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.418334977207902e-06,
+      "loss": 0.1054,
+      "step": 4464
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.418201690080773e-06,
+      "loss": 0.1492,
+      "step": 4465
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.418068402953644e-06,
+      "loss": 0.0866,
+      "step": 4466
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.417935115826515e-06,
+      "loss": 0.1427,
+      "step": 4467
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.417801828699386e-06,
+      "loss": 0.2002,
+      "step": 4468
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.417668541572255e-06,
+      "loss": 0.1099,
+      "step": 4469
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.417535254445126e-06,
+      "loss": 0.0831,
+      "step": 4470
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.417401967317997e-06,
+      "loss": 0.0931,
+      "step": 4471
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.417268680190868e-06,
+      "loss": 0.1112,
+      "step": 4472
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.41713539306374e-06,
+      "loss": 0.0957,
+      "step": 4473
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.417002105936608e-06,
+      "loss": 0.1541,
+      "step": 4474
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.41686881880948e-06,
+      "loss": 0.1044,
+      "step": 4475
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.41673553168235e-06,
+      "loss": 0.0993,
+      "step": 4476
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.416602244555221e-06,
+      "loss": 0.0842,
+      "step": 4477
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.416468957428092e-06,
+      "loss": 0.1129,
+      "step": 4478
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.416335670300963e-06,
+      "loss": 0.1164,
+      "step": 4479
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.416202383173833e-06,
+      "loss": 0.1517,
+      "step": 4480
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.416069096046705e-06,
+      "loss": 0.0929,
+      "step": 4481
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.415935808919575e-06,
+      "loss": 0.0692,
+      "step": 4482
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.415802521792446e-06,
+      "loss": 0.0861,
+      "step": 4483
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.415669234665317e-06,
+      "loss": 0.0906,
+      "step": 4484
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.415535947538188e-06,
+      "loss": 0.0764,
+      "step": 4485
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.415402660411059e-06,
+      "loss": 0.1149,
+      "step": 4486
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.41526937328393e-06,
+      "loss": 0.1647,
+      "step": 4487
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.415136086156799e-06,
+      "loss": 0.0837,
+      "step": 4488
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.41500279902967e-06,
+      "loss": 0.1102,
+      "step": 4489
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.414869511902541e-06,
+      "loss": 0.098,
+      "step": 4490
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.414736224775412e-06,
+      "loss": 0.1719,
+      "step": 4491
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.414602937648283e-06,
+      "loss": 0.0999,
+      "step": 4492
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.414469650521152e-06,
+      "loss": 0.123,
+      "step": 4493
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.414336363394025e-06,
+      "loss": 0.0923,
+      "step": 4494
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.414203076266894e-06,
+      "loss": 0.1432,
+      "step": 4495
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.414069789139765e-06,
+      "loss": 0.1045,
+      "step": 4496
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.413936502012636e-06,
+      "loss": 0.074,
+      "step": 4497
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.413803214885507e-06,
+      "loss": 0.0982,
+      "step": 4498
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.413669927758378e-06,
+      "loss": 0.105,
+      "step": 4499
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.41353664063125e-06,
+      "loss": 0.1075,
+      "step": 4500
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.413403353504119e-06,
+      "loss": 0.1337,
+      "step": 4501
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.413270066376991e-06,
+      "loss": 0.137,
+      "step": 4502
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.41313677924986e-06,
+      "loss": 0.1239,
+      "step": 4503
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.413003492122732e-06,
+      "loss": 0.1619,
+      "step": 4504
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.412870204995603e-06,
+      "loss": 0.1076,
+      "step": 4505
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.412736917868474e-06,
+      "loss": 0.0933,
+      "step": 4506
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.412603630741345e-06,
+      "loss": 0.1102,
+      "step": 4507
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.412470343614214e-06,
+      "loss": 0.1067,
+      "step": 4508
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.412337056487085e-06,
+      "loss": 0.1169,
+      "step": 4509
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.412203769359956e-06,
+      "loss": 0.158,
+      "step": 4510
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.412070482232827e-06,
+      "loss": 0.1012,
+      "step": 4511
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.411937195105698e-06,
+      "loss": 0.0811,
+      "step": 4512
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.411803907978569e-06,
+      "loss": 0.1241,
+      "step": 4513
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.411670620851438e-06,
+      "loss": 0.1184,
+      "step": 4514
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.411537333724309e-06,
+      "loss": 0.1343,
+      "step": 4515
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.41140404659718e-06,
+      "loss": 0.1024,
+      "step": 4516
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.411270759470051e-06,
+      "loss": 0.0849,
+      "step": 4517
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.411137472342922e-06,
+      "loss": 0.0701,
+      "step": 4518
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.411004185215793e-06,
+      "loss": 0.1187,
+      "step": 4519
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.410870898088662e-06,
+      "loss": 0.1444,
+      "step": 4520
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.410737610961535e-06,
+      "loss": 0.1239,
+      "step": 4521
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.410604323834404e-06,
+      "loss": 0.1497,
+      "step": 4522
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.410471036707275e-06,
+      "loss": 0.1279,
+      "step": 4523
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.410337749580146e-06,
+      "loss": 0.0719,
+      "step": 4524
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.410204462453017e-06,
+      "loss": 0.1458,
+      "step": 4525
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.410071175325888e-06,
+      "loss": 0.1025,
+      "step": 4526
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.40993788819876e-06,
+      "loss": 0.1532,
+      "step": 4527
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.409804601071629e-06,
+      "loss": 0.0844,
+      "step": 4528
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.4096713139445e-06,
+      "loss": 0.1208,
+      "step": 4529
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.40953802681737e-06,
+      "loss": 0.1466,
+      "step": 4530
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.409404739690242e-06,
+      "loss": 0.086,
+      "step": 4531
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.409271452563113e-06,
+      "loss": 0.1276,
+      "step": 4532
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.409138165435982e-06,
+      "loss": 0.1354,
+      "step": 4533
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.409004878308855e-06,
+      "loss": 0.1044,
+      "step": 4534
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.408871591181724e-06,
+      "loss": 0.1052,
+      "step": 4535
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.408738304054595e-06,
+      "loss": 0.1046,
+      "step": 4536
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.408605016927466e-06,
+      "loss": 0.1366,
+      "step": 4537
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.408471729800337e-06,
+      "loss": 0.1069,
+      "step": 4538
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.408338442673208e-06,
+      "loss": 0.1458,
+      "step": 4539
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.408205155546079e-06,
+      "loss": 0.0935,
+      "step": 4540
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.408071868418948e-06,
+      "loss": 0.1234,
+      "step": 4541
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.40793858129182e-06,
+      "loss": 0.0929,
+      "step": 4542
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.40780529416469e-06,
+      "loss": 0.1545,
+      "step": 4543
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.407672007037561e-06,
+      "loss": 0.0768,
+      "step": 4544
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.407538719910432e-06,
+      "loss": 0.1431,
+      "step": 4545
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.407405432783303e-06,
+      "loss": 0.1356,
+      "step": 4546
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.407272145656172e-06,
+      "loss": 0.0916,
+      "step": 4547
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.407138858529043e-06,
+      "loss": 0.0882,
+      "step": 4548
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.407005571401914e-06,
+      "loss": 0.0962,
+      "step": 4549
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.406872284274785e-06,
+      "loss": 0.1004,
+      "step": 4550
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.406738997147656e-06,
+      "loss": 0.0941,
+      "step": 4551
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.406605710020526e-06,
+      "loss": 0.1208,
+      "step": 4552
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.406472422893398e-06,
+      "loss": 0.1233,
+      "step": 4553
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.406339135766268e-06,
+      "loss": 0.1023,
+      "step": 4554
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.406205848639139e-06,
+      "loss": 0.0928,
+      "step": 4555
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.40607256151201e-06,
+      "loss": 0.115,
+      "step": 4556
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.40593927438488e-06,
+      "loss": 0.1146,
+      "step": 4557
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.405805987257752e-06,
+      "loss": 0.0792,
+      "step": 4558
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.405672700130623e-06,
+      "loss": 0.085,
+      "step": 4559
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.405539413003492e-06,
+      "loss": 0.1245,
+      "step": 4560
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.405406125876365e-06,
+      "loss": 0.1283,
+      "step": 4561
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.405272838749234e-06,
+      "loss": 0.0811,
+      "step": 4562
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.405139551622105e-06,
+      "loss": 0.1473,
+      "step": 4563
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.405006264494976e-06,
+      "loss": 0.1145,
+      "step": 4564
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.404872977367847e-06,
+      "loss": 0.1017,
+      "step": 4565
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.404739690240718e-06,
+      "loss": 0.0977,
+      "step": 4566
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.404606403113587e-06,
+      "loss": 0.1116,
+      "step": 4567
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.404473115986458e-06,
+      "loss": 0.0936,
+      "step": 4568
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.40433982885933e-06,
+      "loss": 0.0833,
+      "step": 4569
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.4042065417322e-06,
+      "loss": 0.1365,
+      "step": 4570
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.404073254605071e-06,
+      "loss": 0.0951,
+      "step": 4571
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.403939967477942e-06,
+      "loss": 0.1021,
+      "step": 4572
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.403806680350812e-06,
+      "loss": 0.118,
+      "step": 4573
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.403673393223684e-06,
+      "loss": 0.0925,
+      "step": 4574
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.403540106096554e-06,
+      "loss": 0.0833,
+      "step": 4575
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.403406818969425e-06,
+      "loss": 0.1452,
+      "step": 4576
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.403273531842296e-06,
+      "loss": 0.1173,
+      "step": 4577
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.403140244715167e-06,
+      "loss": 0.1162,
+      "step": 4578
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.403006957588038e-06,
+      "loss": 0.1185,
+      "step": 4579
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.402873670460909e-06,
+      "loss": 0.0844,
+      "step": 4580
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.402740383333778e-06,
+      "loss": 0.1364,
+      "step": 4581
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.402607096206649e-06,
+      "loss": 0.0871,
+      "step": 4582
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.40247380907952e-06,
+      "loss": 0.108,
+      "step": 4583
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.40234052195239e-06,
+      "loss": 0.1164,
+      "step": 4584
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.402207234825262e-06,
+      "loss": 0.1433,
+      "step": 4585
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.402073947698131e-06,
+      "loss": 0.1296,
+      "step": 4586
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.401940660571002e-06,
+      "loss": 0.1182,
+      "step": 4587
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.401807373443873e-06,
+      "loss": 0.0812,
+      "step": 4588
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.401674086316744e-06,
+      "loss": 0.0941,
+      "step": 4589
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.401540799189615e-06,
+      "loss": 0.1108,
+      "step": 4590
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.401407512062486e-06,
+      "loss": 0.1721,
+      "step": 4591
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.401274224935355e-06,
+      "loss": 0.099,
+      "step": 4592
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.401140937808228e-06,
+      "loss": 0.1152,
+      "step": 4593
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.401007650681097e-06,
+      "loss": 0.1169,
+      "step": 4594
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.400874363553968e-06,
+      "loss": 0.1007,
+      "step": 4595
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.40074107642684e-06,
+      "loss": 0.1319,
+      "step": 4596
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.40060778929971e-06,
+      "loss": 0.0942,
+      "step": 4597
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.400474502172581e-06,
+      "loss": 0.1067,
+      "step": 4598
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.400341215045452e-06,
+      "loss": 0.0945,
+      "step": 4599
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.400207927918322e-06,
+      "loss": 0.0925,
+      "step": 4600
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.400074640791194e-06,
+      "loss": 0.129,
+      "step": 4601
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.399941353664064e-06,
+      "loss": 0.1699,
+      "step": 4602
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.399808066536935e-06,
+      "loss": 0.1295,
+      "step": 4603
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.399674779409806e-06,
+      "loss": 0.1411,
+      "step": 4604
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.399541492282677e-06,
+      "loss": 0.1025,
+      "step": 4605
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.399408205155548e-06,
+      "loss": 0.0908,
+      "step": 4606
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.399274918028417e-06,
+      "loss": 0.1117,
+      "step": 4607
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.399141630901288e-06,
+      "loss": 0.1541,
+      "step": 4608
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.399008343774159e-06,
+      "loss": 0.0832,
+      "step": 4609
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.39887505664703e-06,
+      "loss": 0.1024,
+      "step": 4610
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.398741769519901e-06,
+      "loss": 0.122,
+      "step": 4611
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.398608482392772e-06,
+      "loss": 0.1198,
+      "step": 4612
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.398475195265641e-06,
+      "loss": 0.1073,
+      "step": 4613
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.398341908138512e-06,
+      "loss": 0.0731,
+      "step": 4614
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.398208621011383e-06,
+      "loss": 0.115,
+      "step": 4615
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.398075333884254e-06,
+      "loss": 0.1041,
+      "step": 4616
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.397942046757125e-06,
+      "loss": 0.09,
+      "step": 4617
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.397808759629996e-06,
+      "loss": 0.1395,
+      "step": 4618
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.397675472502865e-06,
+      "loss": 0.117,
+      "step": 4619
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.397542185375738e-06,
+      "loss": 0.0953,
+      "step": 4620
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.397408898248607e-06,
+      "loss": 0.1129,
+      "step": 4621
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.397275611121478e-06,
+      "loss": 0.0982,
+      "step": 4622
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.39714232399435e-06,
+      "loss": 0.1257,
+      "step": 4623
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.39700903686722e-06,
+      "loss": 0.1224,
+      "step": 4624
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.396875749740091e-06,
+      "loss": 0.117,
+      "step": 4625
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.39674246261296e-06,
+      "loss": 0.1647,
+      "step": 4626
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.396609175485832e-06,
+      "loss": 0.0971,
+      "step": 4627
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.396475888358703e-06,
+      "loss": 0.1703,
+      "step": 4628
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.396342601231574e-06,
+      "loss": 0.1123,
+      "step": 4629
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.396209314104445e-06,
+      "loss": 0.0726,
+      "step": 4630
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.396076026977316e-06,
+      "loss": 0.1078,
+      "step": 4631
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.395942739850185e-06,
+      "loss": 0.0957,
+      "step": 4632
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.395809452723058e-06,
+      "loss": 0.0979,
+      "step": 4633
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.395676165595927e-06,
+      "loss": 0.0793,
+      "step": 4634
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.395542878468798e-06,
+      "loss": 0.1275,
+      "step": 4635
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.395409591341669e-06,
+      "loss": 0.1174,
+      "step": 4636
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.39527630421454e-06,
+      "loss": 0.114,
+      "step": 4637
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.395143017087411e-06,
+      "loss": 0.1569,
+      "step": 4638
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.395009729960282e-06,
+      "loss": 0.1034,
+      "step": 4639
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.394876442833151e-06,
+      "loss": 0.108,
+      "step": 4640
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.394743155706024e-06,
+      "loss": 0.1044,
+      "step": 4641
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.394609868578893e-06,
+      "loss": 0.1265,
+      "step": 4642
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.394476581451764e-06,
+      "loss": 0.1223,
+      "step": 4643
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.394343294324635e-06,
+      "loss": 0.1362,
+      "step": 4644
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.394210007197505e-06,
+      "loss": 0.0952,
+      "step": 4645
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.394076720070377e-06,
+      "loss": 0.1094,
+      "step": 4646
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.393943432943247e-06,
+      "loss": 0.1292,
+      "step": 4647
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.393810145816118e-06,
+      "loss": 0.111,
+      "step": 4648
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.393676858688989e-06,
+      "loss": 0.1326,
+      "step": 4649
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.39354357156186e-06,
+      "loss": 0.1097,
+      "step": 4650
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.39341028443473e-06,
+      "loss": 0.1172,
+      "step": 4651
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.393276997307602e-06,
+      "loss": 0.1086,
+      "step": 4652
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.39314371018047e-06,
+      "loss": 0.1017,
+      "step": 4653
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.393010423053342e-06,
+      "loss": 0.1169,
+      "step": 4654
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.392877135926213e-06,
+      "loss": 0.0838,
+      "step": 4655
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.392743848799084e-06,
+      "loss": 0.1351,
+      "step": 4656
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.392610561671955e-06,
+      "loss": 0.1376,
+      "step": 4657
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.392477274544826e-06,
+      "loss": 0.1026,
+      "step": 4658
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.392343987417695e-06,
+      "loss": 0.0756,
+      "step": 4659
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.392210700290568e-06,
+      "loss": 0.113,
+      "step": 4660
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.392077413163437e-06,
+      "loss": 0.0855,
+      "step": 4661
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.391944126036308e-06,
+      "loss": 0.1107,
+      "step": 4662
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.391810838909179e-06,
+      "loss": 0.1038,
+      "step": 4663
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.39167755178205e-06,
+      "loss": 0.1203,
+      "step": 4664
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.391544264654921e-06,
+      "loss": 0.0826,
+      "step": 4665
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.39141097752779e-06,
+      "loss": 0.0881,
+      "step": 4666
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.391277690400661e-06,
+      "loss": 0.1175,
+      "step": 4667
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.391144403273532e-06,
+      "loss": 0.1065,
+      "step": 4668
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.391011116146403e-06,
+      "loss": 0.1224,
+      "step": 4669
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.390877829019274e-06,
+      "loss": 0.0956,
+      "step": 4670
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.390744541892145e-06,
+      "loss": 0.0851,
+      "step": 4671
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.390611254765015e-06,
+      "loss": 0.0862,
+      "step": 4672
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.390477967637887e-06,
+      "loss": 0.1195,
+      "step": 4673
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.390344680510757e-06,
+      "loss": 0.1465,
+      "step": 4674
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.390211393383628e-06,
+      "loss": 0.1132,
+      "step": 4675
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.390078106256499e-06,
+      "loss": 0.1353,
+      "step": 4676
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.38994481912937e-06,
+      "loss": 0.1391,
+      "step": 4677
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.38981153200224e-06,
+      "loss": 0.0825,
+      "step": 4678
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.389678244875112e-06,
+      "loss": 0.1246,
+      "step": 4679
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.389544957747981e-06,
+      "loss": 0.1294,
+      "step": 4680
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.389411670620852e-06,
+      "loss": 0.1422,
+      "step": 4681
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.389278383493723e-06,
+      "loss": 0.0885,
+      "step": 4682
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.389145096366594e-06,
+      "loss": 0.1342,
+      "step": 4683
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.389011809239465e-06,
+      "loss": 0.0886,
+      "step": 4684
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.388878522112334e-06,
+      "loss": 0.105,
+      "step": 4685
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.388745234985205e-06,
+      "loss": 0.1025,
+      "step": 4686
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.388611947858076e-06,
+      "loss": 0.1152,
+      "step": 4687
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.388478660730947e-06,
+      "loss": 0.1354,
+      "step": 4688
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.388345373603818e-06,
+      "loss": 0.095,
+      "step": 4689
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.388212086476689e-06,
+      "loss": 0.1386,
+      "step": 4690
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.388078799349558e-06,
+      "loss": 0.0822,
+      "step": 4691
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.387945512222431e-06,
+      "loss": 0.0925,
+      "step": 4692
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.3878122250953e-06,
+      "loss": 0.1315,
+      "step": 4693
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.387678937968171e-06,
+      "loss": 0.1336,
+      "step": 4694
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.387545650841042e-06,
+      "loss": 0.074,
+      "step": 4695
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.387412363713913e-06,
+      "loss": 0.136,
+      "step": 4696
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.387279076586784e-06,
+      "loss": 0.0927,
+      "step": 4697
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.387145789459655e-06,
+      "loss": 0.102,
+      "step": 4698
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.387012502332525e-06,
+      "loss": 0.1138,
+      "step": 4699
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.386879215205397e-06,
+      "loss": 0.1125,
+      "step": 4700
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.386745928078267e-06,
+      "loss": 0.1016,
+      "step": 4701
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.386612640951138e-06,
+      "loss": 0.1669,
+      "step": 4702
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.386479353824009e-06,
+      "loss": 0.1319,
+      "step": 4703
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.386346066696878e-06,
+      "loss": 0.1468,
+      "step": 4704
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.38621277956975e-06,
+      "loss": 0.0777,
+      "step": 4705
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.38607949244262e-06,
+      "loss": 0.1643,
+      "step": 4706
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.385946205315491e-06,
+      "loss": 0.1245,
+      "step": 4707
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.385812918188362e-06,
+      "loss": 0.106,
+      "step": 4708
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.385679631061233e-06,
+      "loss": 0.1074,
+      "step": 4709
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.385546343934104e-06,
+      "loss": 0.0859,
+      "step": 4710
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.385413056806975e-06,
+      "loss": 0.1338,
+      "step": 4711
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.385279769679844e-06,
+      "loss": 0.1148,
+      "step": 4712
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.385146482552717e-06,
+      "loss": 0.0783,
+      "step": 4713
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.385013195425586e-06,
+      "loss": 0.0867,
+      "step": 4714
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.384879908298457e-06,
+      "loss": 0.1324,
+      "step": 4715
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.384746621171328e-06,
+      "loss": 0.1161,
+      "step": 4716
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.3846133340442e-06,
+      "loss": 0.1113,
+      "step": 4717
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.38448004691707e-06,
+      "loss": 0.1129,
+      "step": 4718
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.384346759789941e-06,
+      "loss": 0.0717,
+      "step": 4719
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.38421347266281e-06,
+      "loss": 0.0627,
+      "step": 4720
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.384080185535682e-06,
+      "loss": 0.076,
+      "step": 4721
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.383946898408553e-06,
+      "loss": 0.1332,
+      "step": 4722
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.383813611281424e-06,
+      "loss": 0.1145,
+      "step": 4723
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.383680324154294e-06,
+      "loss": 0.1583,
+      "step": 4724
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.383547037027164e-06,
+      "loss": 0.1034,
+      "step": 4725
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.383413749900035e-06,
+      "loss": 0.0888,
+      "step": 4726
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.383280462772906e-06,
+      "loss": 0.0981,
+      "step": 4727
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.383147175645777e-06,
+      "loss": 0.1238,
+      "step": 4728
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.383013888518648e-06,
+      "loss": 0.1148,
+      "step": 4729
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.382880601391519e-06,
+      "loss": 0.137,
+      "step": 4730
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.382747314264388e-06,
+      "loss": 0.1386,
+      "step": 4731
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.38261402713726e-06,
+      "loss": 0.132,
+      "step": 4732
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.38248074001013e-06,
+      "loss": 0.0945,
+      "step": 4733
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.382347452883001e-06,
+      "loss": 0.098,
+      "step": 4734
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.382214165755872e-06,
+      "loss": 0.1628,
+      "step": 4735
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.382080878628743e-06,
+      "loss": 0.1409,
+      "step": 4736
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.381947591501614e-06,
+      "loss": 0.0997,
+      "step": 4737
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.381814304374485e-06,
+      "loss": 0.1371,
+      "step": 4738
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.381681017247354e-06,
+      "loss": 0.0808,
+      "step": 4739
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.381547730120227e-06,
+      "loss": 0.1213,
+      "step": 4740
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.381414442993096e-06,
+      "loss": 0.0935,
+      "step": 4741
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.381281155865967e-06,
+      "loss": 0.1144,
+      "step": 4742
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.381147868738838e-06,
+      "loss": 0.1679,
+      "step": 4743
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.381014581611708e-06,
+      "loss": 0.1444,
+      "step": 4744
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.38088129448458e-06,
+      "loss": 0.1222,
+      "step": 4745
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.38074800735745e-06,
+      "loss": 0.1141,
+      "step": 4746
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.38061472023032e-06,
+      "loss": 0.0971,
+      "step": 4747
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.380481433103192e-06,
+      "loss": 0.0756,
+      "step": 4748
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.380348145976063e-06,
+      "loss": 0.1216,
+      "step": 4749
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.380214858848934e-06,
+      "loss": 0.1174,
+      "step": 4750
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.380081571721805e-06,
+      "loss": 0.0829,
+      "step": 4751
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.379948284594674e-06,
+      "loss": 0.0829,
+      "step": 4752
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.379814997467545e-06,
+      "loss": 0.1015,
+      "step": 4753
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.379681710340416e-06,
+      "loss": 0.1255,
+      "step": 4754
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.379548423213287e-06,
+      "loss": 0.1301,
+      "step": 4755
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.379415136086158e-06,
+      "loss": 0.1246,
+      "step": 4756
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.379281848959029e-06,
+      "loss": 0.1084,
+      "step": 4757
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.379148561831898e-06,
+      "loss": 0.1336,
+      "step": 4758
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.37901527470477e-06,
+      "loss": 0.126,
+      "step": 4759
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.37888198757764e-06,
+      "loss": 0.1093,
+      "step": 4760
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.378748700450511e-06,
+      "loss": 0.1273,
+      "step": 4761
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.378615413323382e-06,
+      "loss": 0.1199,
+      "step": 4762
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.378482126196251e-06,
+      "loss": 0.1076,
+      "step": 4763
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.378348839069124e-06,
+      "loss": 0.0865,
+      "step": 4764
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.378215551941993e-06,
+      "loss": 0.1059,
+      "step": 4765
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.378082264814864e-06,
+      "loss": 0.1151,
+      "step": 4766
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.377948977687735e-06,
+      "loss": 0.1228,
+      "step": 4767
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.377815690560606e-06,
+      "loss": 0.1904,
+      "step": 4768
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.377682403433477e-06,
+      "loss": 0.109,
+      "step": 4769
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.377549116306348e-06,
+      "loss": 0.1219,
+      "step": 4770
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.377415829179218e-06,
+      "loss": 0.1178,
+      "step": 4771
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.37728254205209e-06,
+      "loss": 0.1259,
+      "step": 4772
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.37714925492496e-06,
+      "loss": 0.1035,
+      "step": 4773
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.37701596779783e-06,
+      "loss": 0.0903,
+      "step": 4774
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.376882680670702e-06,
+      "loss": 0.091,
+      "step": 4775
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.376749393543573e-06,
+      "loss": 0.0873,
+      "step": 4776
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.376616106416444e-06,
+      "loss": 0.1145,
+      "step": 4777
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.376482819289315e-06,
+      "loss": 0.1196,
+      "step": 4778
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.376349532162184e-06,
+      "loss": 0.1138,
+      "step": 4779
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.376216245035055e-06,
+      "loss": 0.095,
+      "step": 4780
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.376082957907926e-06,
+      "loss": 0.1059,
+      "step": 4781
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.375949670780797e-06,
+      "loss": 0.1074,
+      "step": 4782
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.375816383653668e-06,
+      "loss": 0.1122,
+      "step": 4783
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.375683096526537e-06,
+      "loss": 0.104,
+      "step": 4784
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.375549809399408e-06,
+      "loss": 0.1091,
+      "step": 4785
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.37541652227228e-06,
+      "loss": 0.1149,
+      "step": 4786
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.37528323514515e-06,
+      "loss": 0.1403,
+      "step": 4787
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.375149948018021e-06,
+      "loss": 0.1297,
+      "step": 4788
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.375016660890892e-06,
+      "loss": 0.092,
+      "step": 4789
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.374883373763762e-06,
+      "loss": 0.1189,
+      "step": 4790
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.374750086636634e-06,
+      "loss": 0.1721,
+      "step": 4791
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.374616799509504e-06,
+      "loss": 0.0842,
+      "step": 4792
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.374483512382374e-06,
+      "loss": 0.1219,
+      "step": 4793
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.374350225255245e-06,
+      "loss": 0.1413,
+      "step": 4794
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.374216938128116e-06,
+      "loss": 0.0655,
+      "step": 4795
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.374083651000987e-06,
+      "loss": 0.1524,
+      "step": 4796
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.373950363873858e-06,
+      "loss": 0.1184,
+      "step": 4797
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.373817076746728e-06,
+      "loss": 0.1225,
+      "step": 4798
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.3736837896196e-06,
+      "loss": 0.1291,
+      "step": 4799
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.37355050249247e-06,
+      "loss": 0.1062,
+      "step": 4800
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.37341721536534e-06,
+      "loss": 0.1186,
+      "step": 4801
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.373283928238212e-06,
+      "loss": 0.1362,
+      "step": 4802
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.373150641111081e-06,
+      "loss": 0.1153,
+      "step": 4803
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.373017353983954e-06,
+      "loss": 0.108,
+      "step": 4804
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.372884066856823e-06,
+      "loss": 0.1126,
+      "step": 4805
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.372750779729694e-06,
+      "loss": 0.116,
+      "step": 4806
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.372617492602565e-06,
+      "loss": 0.0812,
+      "step": 4807
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.372484205475436e-06,
+      "loss": 0.1047,
+      "step": 4808
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.372350918348307e-06,
+      "loss": 0.1292,
+      "step": 4809
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.372217631221178e-06,
+      "loss": 0.1091,
+      "step": 4810
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.372084344094047e-06,
+      "loss": 0.1692,
+      "step": 4811
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.37195105696692e-06,
+      "loss": 0.1521,
+      "step": 4812
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.37181776983979e-06,
+      "loss": 0.1104,
+      "step": 4813
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.37168448271266e-06,
+      "loss": 0.072,
+      "step": 4814
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.371551195585531e-06,
+      "loss": 0.1198,
+      "step": 4815
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.371417908458402e-06,
+      "loss": 0.089,
+      "step": 4816
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.371284621331273e-06,
+      "loss": 0.1568,
+      "step": 4817
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.371151334204144e-06,
+      "loss": 0.1097,
+      "step": 4818
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.371018047077014e-06,
+      "loss": 0.1125,
+      "step": 4819
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.370884759949885e-06,
+      "loss": 0.0836,
+      "step": 4820
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.370751472822756e-06,
+      "loss": 0.0874,
+      "step": 4821
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.370618185695627e-06,
+      "loss": 0.0917,
+      "step": 4822
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.370484898568498e-06,
+      "loss": 0.1138,
+      "step": 4823
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.370351611441367e-06,
+      "loss": 0.1344,
+      "step": 4824
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.370218324314238e-06,
+      "loss": 0.1335,
+      "step": 4825
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.370085037187109e-06,
+      "loss": 0.1311,
+      "step": 4826
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.36995175005998e-06,
+      "loss": 0.1285,
+      "step": 4827
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.36981846293285e-06,
+      "loss": 0.105,
+      "step": 4828
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.369685175805722e-06,
+      "loss": 0.0854,
+      "step": 4829
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.369551888678591e-06,
+      "loss": 0.0942,
+      "step": 4830
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.369418601551464e-06,
+      "loss": 0.0971,
+      "step": 4831
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.369285314424333e-06,
+      "loss": 0.0829,
+      "step": 4832
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.369152027297204e-06,
+      "loss": 0.0878,
+      "step": 4833
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.369018740170075e-06,
+      "loss": 0.1252,
+      "step": 4834
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.368885453042946e-06,
+      "loss": 0.089,
+      "step": 4835
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.368752165915817e-06,
+      "loss": 0.0864,
+      "step": 4836
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.368618878788688e-06,
+      "loss": 0.114,
+      "step": 4837
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.368485591661557e-06,
+      "loss": 0.0918,
+      "step": 4838
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.36835230453443e-06,
+      "loss": 0.1184,
+      "step": 4839
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.3682190174073e-06,
+      "loss": 0.0823,
+      "step": 4840
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.36808573028017e-06,
+      "loss": 0.1368,
+      "step": 4841
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.367952443153041e-06,
+      "loss": 0.0964,
+      "step": 4842
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.36781915602591e-06,
+      "loss": 0.1285,
+      "step": 4843
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.367685868898783e-06,
+      "loss": 0.0942,
+      "step": 4844
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.367552581771653e-06,
+      "loss": 0.1096,
+      "step": 4845
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.367419294644524e-06,
+      "loss": 0.0749,
+      "step": 4846
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.367286007517395e-06,
+      "loss": 0.1329,
+      "step": 4847
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.367152720390266e-06,
+      "loss": 0.1223,
+      "step": 4848
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.367019433263137e-06,
+      "loss": 0.0889,
+      "step": 4849
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.366886146136008e-06,
+      "loss": 0.1195,
+      "step": 4850
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.366752859008877e-06,
+      "loss": 0.1341,
+      "step": 4851
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.366619571881748e-06,
+      "loss": 0.0862,
+      "step": 4852
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.366486284754619e-06,
+      "loss": 0.1007,
+      "step": 4853
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.36635299762749e-06,
+      "loss": 0.115,
+      "step": 4854
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.366219710500361e-06,
+      "loss": 0.0823,
+      "step": 4855
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.366086423373232e-06,
+      "loss": 0.0966,
+      "step": 4856
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.365953136246101e-06,
+      "loss": 0.1079,
+      "step": 4857
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.365819849118974e-06,
+      "loss": 0.1063,
+      "step": 4858
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.365686561991843e-06,
+      "loss": 0.1483,
+      "step": 4859
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.365553274864714e-06,
+      "loss": 0.1056,
+      "step": 4860
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.365419987737585e-06,
+      "loss": 0.1048,
+      "step": 4861
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.365286700610454e-06,
+      "loss": 0.1414,
+      "step": 4862
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.365153413483327e-06,
+      "loss": 0.098,
+      "step": 4863
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.365020126356196e-06,
+      "loss": 0.1269,
+      "step": 4864
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.364886839229067e-06,
+      "loss": 0.0964,
+      "step": 4865
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.364753552101938e-06,
+      "loss": 0.1164,
+      "step": 4866
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.36462026497481e-06,
+      "loss": 0.121,
+      "step": 4867
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.36448697784768e-06,
+      "loss": 0.1122,
+      "step": 4868
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.364353690720551e-06,
+      "loss": 0.1455,
+      "step": 4869
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.36422040359342e-06,
+      "loss": 0.0931,
+      "step": 4870
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.364087116466293e-06,
+      "loss": 0.1065,
+      "step": 4871
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.363953829339163e-06,
+      "loss": 0.0692,
+      "step": 4872
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.363820542212034e-06,
+      "loss": 0.1132,
+      "step": 4873
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.363687255084905e-06,
+      "loss": 0.1077,
+      "step": 4874
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.363553967957776e-06,
+      "loss": 0.1146,
+      "step": 4875
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.363420680830647e-06,
+      "loss": 0.0889,
+      "step": 4876
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.363287393703518e-06,
+      "loss": 0.0793,
+      "step": 4877
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.363154106576387e-06,
+      "loss": 0.1191,
+      "step": 4878
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.36302081944926e-06,
+      "loss": 0.0982,
+      "step": 4879
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.362887532322129e-06,
+      "loss": 0.1248,
+      "step": 4880
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.362754245195e-06,
+      "loss": 0.1207,
+      "step": 4881
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.362620958067871e-06,
+      "loss": 0.1184,
+      "step": 4882
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.36248767094074e-06,
+      "loss": 0.1359,
+      "step": 4883
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.362354383813613e-06,
+      "loss": 0.0998,
+      "step": 4884
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.362221096686482e-06,
+      "loss": 0.1094,
+      "step": 4885
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.362087809559353e-06,
+      "loss": 0.1091,
+      "step": 4886
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.361954522432224e-06,
+      "loss": 0.1061,
+      "step": 4887
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.361821235305095e-06,
+      "loss": 0.1329,
+      "step": 4888
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.361687948177966e-06,
+      "loss": 0.0866,
+      "step": 4889
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.361554661050837e-06,
+      "loss": 0.1311,
+      "step": 4890
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.361421373923707e-06,
+      "loss": 0.1397,
+      "step": 4891
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.361288086796578e-06,
+      "loss": 0.0827,
+      "step": 4892
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.361154799669449e-06,
+      "loss": 0.1442,
+      "step": 4893
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.36102151254232e-06,
+      "loss": 0.1096,
+      "step": 4894
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.36088822541519e-06,
+      "loss": 0.0784,
+      "step": 4895
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.360754938288062e-06,
+      "loss": 0.0994,
+      "step": 4896
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.36062165116093e-06,
+      "loss": 0.1171,
+      "step": 4897
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.360488364033804e-06,
+      "loss": 0.14,
+      "step": 4898
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.360355076906673e-06,
+      "loss": 0.1525,
+      "step": 4899
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.360221789779544e-06,
+      "loss": 0.1159,
+      "step": 4900
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.360088502652415e-06,
+      "loss": 0.0958,
+      "step": 4901
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.359955215525284e-06,
+      "loss": 0.0904,
+      "step": 4902
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.359821928398157e-06,
+      "loss": 0.1269,
+      "step": 4903
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.359688641271026e-06,
+      "loss": 0.0922,
+      "step": 4904
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.359555354143897e-06,
+      "loss": 0.0891,
+      "step": 4905
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.359422067016768e-06,
+      "loss": 0.1056,
+      "step": 4906
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.359288779889639e-06,
+      "loss": 0.1018,
+      "step": 4907
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.35915549276251e-06,
+      "loss": 0.1068,
+      "step": 4908
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.359022205635381e-06,
+      "loss": 0.104,
+      "step": 4909
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.35888891850825e-06,
+      "loss": 0.1482,
+      "step": 4910
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.358755631381123e-06,
+      "loss": 0.1028,
+      "step": 4911
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.358622344253992e-06,
+      "loss": 0.1345,
+      "step": 4912
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.358489057126863e-06,
+      "loss": 0.0827,
+      "step": 4913
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.358355769999734e-06,
+      "loss": 0.1238,
+      "step": 4914
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.358222482872605e-06,
+      "loss": 0.1439,
+      "step": 4915
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.358089195745476e-06,
+      "loss": 0.1081,
+      "step": 4916
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.357955908618347e-06,
+      "loss": 0.1277,
+      "step": 4917
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.357822621491217e-06,
+      "loss": 0.1467,
+      "step": 4918
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.357689334364088e-06,
+      "loss": 0.1241,
+      "step": 4919
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.357556047236959e-06,
+      "loss": 0.1083,
+      "step": 4920
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.35742276010983e-06,
+      "loss": 0.1717,
+      "step": 4921
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.3572894729827e-06,
+      "loss": 0.1092,
+      "step": 4922
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.35715618585557e-06,
+      "loss": 0.0819,
+      "step": 4923
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.357022898728441e-06,
+      "loss": 0.1281,
+      "step": 4924
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.356889611601312e-06,
+      "loss": 0.1148,
+      "step": 4925
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.356756324474183e-06,
+      "loss": 0.1329,
+      "step": 4926
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.356623037347054e-06,
+      "loss": 0.133,
+      "step": 4927
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.356489750219925e-06,
+      "loss": 0.1589,
+      "step": 4928
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.356356463092794e-06,
+      "loss": 0.1036,
+      "step": 4929
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.356223175965667e-06,
+      "loss": 0.0942,
+      "step": 4930
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.356089888838536e-06,
+      "loss": 0.0809,
+      "step": 4931
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.355956601711407e-06,
+      "loss": 0.0996,
+      "step": 4932
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.355823314584278e-06,
+      "loss": 0.1014,
+      "step": 4933
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.35569002745715e-06,
+      "loss": 0.1248,
+      "step": 4934
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.35555674033002e-06,
+      "loss": 0.1112,
+      "step": 4935
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.355423453202891e-06,
+      "loss": 0.1393,
+      "step": 4936
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.35529016607576e-06,
+      "loss": 0.0892,
+      "step": 4937
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.355156878948633e-06,
+      "loss": 0.1073,
+      "step": 4938
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.355023591821502e-06,
+      "loss": 0.1123,
+      "step": 4939
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.354890304694373e-06,
+      "loss": 0.1461,
+      "step": 4940
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.354757017567244e-06,
+      "loss": 0.1035,
+      "step": 4941
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.354623730440114e-06,
+      "loss": 0.1475,
+      "step": 4942
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.354490443312986e-06,
+      "loss": 0.1641,
+      "step": 4943
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.354357156185856e-06,
+      "loss": 0.1175,
+      "step": 4944
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.354223869058727e-06,
+      "loss": 0.1108,
+      "step": 4945
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.354090581931598e-06,
+      "loss": 0.125,
+      "step": 4946
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.353957294804469e-06,
+      "loss": 0.1254,
+      "step": 4947
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.35382400767734e-06,
+      "loss": 0.146,
+      "step": 4948
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.35369072055021e-06,
+      "loss": 0.128,
+      "step": 4949
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.35355743342308e-06,
+      "loss": 0.0764,
+      "step": 4950
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.353424146295953e-06,
+      "loss": 0.1078,
+      "step": 4951
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.353290859168822e-06,
+      "loss": 0.1198,
+      "step": 4952
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.353157572041693e-06,
+      "loss": 0.0837,
+      "step": 4953
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.353024284914564e-06,
+      "loss": 0.113,
+      "step": 4954
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.352890997787435e-06,
+      "loss": 0.122,
+      "step": 4955
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.352757710660306e-06,
+      "loss": 0.1014,
+      "step": 4956
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.352624423533177e-06,
+      "loss": 0.091,
+      "step": 4957
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.352491136406046e-06,
+      "loss": 0.1292,
+      "step": 4958
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.352357849278917e-06,
+      "loss": 0.1276,
+      "step": 4959
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.352224562151788e-06,
+      "loss": 0.1273,
+      "step": 4960
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.35209127502466e-06,
+      "loss": 0.1216,
+      "step": 4961
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.35195798789753e-06,
+      "loss": 0.0942,
+      "step": 4962
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.3518247007704e-06,
+      "loss": 0.1034,
+      "step": 4963
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.35169141364327e-06,
+      "loss": 0.0712,
+      "step": 4964
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.351558126516142e-06,
+      "loss": 0.1607,
+      "step": 4965
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.351424839389013e-06,
+      "loss": 0.1217,
+      "step": 4966
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.351291552261884e-06,
+      "loss": 0.1174,
+      "step": 4967
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.351158265134755e-06,
+      "loss": 0.1061,
+      "step": 4968
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.351024978007624e-06,
+      "loss": 0.1189,
+      "step": 4969
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.350891690880497e-06,
+      "loss": 0.1091,
+      "step": 4970
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.350758403753366e-06,
+      "loss": 0.067,
+      "step": 4971
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.350625116626237e-06,
+      "loss": 0.1014,
+      "step": 4972
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.350491829499108e-06,
+      "loss": 0.1013,
+      "step": 4973
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.350358542371979e-06,
+      "loss": 0.1367,
+      "step": 4974
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.35022525524485e-06,
+      "loss": 0.1649,
+      "step": 4975
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.35009196811772e-06,
+      "loss": 0.099,
+      "step": 4976
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.34995868099059e-06,
+      "loss": 0.0863,
+      "step": 4977
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.349825393863461e-06,
+      "loss": 0.0991,
+      "step": 4978
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.349692106736332e-06,
+      "loss": 0.1042,
+      "step": 4979
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.349558819609203e-06,
+      "loss": 0.113,
+      "step": 4980
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.349425532482074e-06,
+      "loss": 0.1287,
+      "step": 4981
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.349292245354943e-06,
+      "loss": 0.1366,
+      "step": 4982
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.349158958227816e-06,
+      "loss": 0.1065,
+      "step": 4983
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.349025671100685e-06,
+      "loss": 0.1439,
+      "step": 4984
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.348892383973556e-06,
+      "loss": 0.1249,
+      "step": 4985
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.348759096846427e-06,
+      "loss": 0.1127,
+      "step": 4986
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.348625809719298e-06,
+      "loss": 0.1062,
+      "step": 4987
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.34849252259217e-06,
+      "loss": 0.1394,
+      "step": 4988
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.34835923546504e-06,
+      "loss": 0.1157,
+      "step": 4989
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.34822594833791e-06,
+      "loss": 0.0848,
+      "step": 4990
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.34809266121078e-06,
+      "loss": 0.0827,
+      "step": 4991
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.347959374083652e-06,
+      "loss": 0.136,
+      "step": 4992
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.347826086956523e-06,
+      "loss": 0.0915,
+      "step": 4993
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.347692799829394e-06,
+      "loss": 0.1347,
+      "step": 4994
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.347559512702265e-06,
+      "loss": 0.0981,
+      "step": 4995
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.347426225575134e-06,
+      "loss": 0.1071,
+      "step": 4996
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.347292938448007e-06,
+      "loss": 0.0968,
+      "step": 4997
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.347159651320876e-06,
+      "loss": 0.115,
+      "step": 4998
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.347026364193747e-06,
+      "loss": 0.0835,
+      "step": 4999
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.346893077066618e-06,
+      "loss": 0.1573,
+      "step": 5000
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.346759789939487e-06,
+      "loss": 0.1129,
+      "step": 5001
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.34662650281236e-06,
+      "loss": 0.1026,
+      "step": 5002
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.34649321568523e-06,
+      "loss": 0.137,
+      "step": 5003
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.3463599285581e-06,
+      "loss": 0.0749,
+      "step": 5004
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.346226641430971e-06,
+      "loss": 0.0794,
+      "step": 5005
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.346093354303842e-06,
+      "loss": 0.1302,
+      "step": 5006
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.345960067176713e-06,
+      "loss": 0.1107,
+      "step": 5007
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.345826780049584e-06,
+      "loss": 0.1324,
+      "step": 5008
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.345693492922453e-06,
+      "loss": 0.1287,
+      "step": 5009
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.345560205795326e-06,
+      "loss": 0.1852,
+      "step": 5010
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.345426918668195e-06,
+      "loss": 0.0966,
+      "step": 5011
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.345293631541066e-06,
+      "loss": 0.1636,
+      "step": 5012
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.345160344413937e-06,
+      "loss": 0.0892,
+      "step": 5013
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.345027057286808e-06,
+      "loss": 0.1456,
+      "step": 5014
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.34489377015968e-06,
+      "loss": 0.0756,
+      "step": 5015
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.34476048303255e-06,
+      "loss": 0.1161,
+      "step": 5016
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.34462719590542e-06,
+      "loss": 0.1165,
+      "step": 5017
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.34449390877829e-06,
+      "loss": 0.1325,
+      "step": 5018
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.344360621651162e-06,
+      "loss": 0.1287,
+      "step": 5019
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.344227334524033e-06,
+      "loss": 0.1284,
+      "step": 5020
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.344094047396904e-06,
+      "loss": 0.0893,
+      "step": 5021
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.343960760269773e-06,
+      "loss": 0.1359,
+      "step": 5022
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.343827473142644e-06,
+      "loss": 0.0707,
+      "step": 5023
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.343694186015515e-06,
+      "loss": 0.0762,
+      "step": 5024
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.343560898888386e-06,
+      "loss": 0.1738,
+      "step": 5025
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.343427611761257e-06,
+      "loss": 0.1011,
+      "step": 5026
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.343294324634128e-06,
+      "loss": 0.122,
+      "step": 5027
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.343161037506997e-06,
+      "loss": 0.1097,
+      "step": 5028
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.34302775037987e-06,
+      "loss": 0.1004,
+      "step": 5029
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.34289446325274e-06,
+      "loss": 0.143,
+      "step": 5030
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.34276117612561e-06,
+      "loss": 0.1663,
+      "step": 5031
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.342627888998481e-06,
+      "loss": 0.1166,
+      "step": 5032
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.342494601871352e-06,
+      "loss": 0.152,
+      "step": 5033
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.342361314744223e-06,
+      "loss": 0.0766,
+      "step": 5034
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.342228027617094e-06,
+      "loss": 0.1065,
+      "step": 5035
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.342094740489964e-06,
+      "loss": 0.0916,
+      "step": 5036
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.341961453362835e-06,
+      "loss": 0.0961,
+      "step": 5037
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.341828166235706e-06,
+      "loss": 0.0795,
+      "step": 5038
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.341694879108577e-06,
+      "loss": 0.0962,
+      "step": 5039
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.341561591981448e-06,
+      "loss": 0.084,
+      "step": 5040
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.341428304854317e-06,
+      "loss": 0.0817,
+      "step": 5041
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.34129501772719e-06,
+      "loss": 0.1109,
+      "step": 5042
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.341161730600059e-06,
+      "loss": 0.1045,
+      "step": 5043
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.34102844347293e-06,
+      "loss": 0.0795,
+      "step": 5044
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.3408951563458e-06,
+      "loss": 0.1107,
+      "step": 5045
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.340761869218672e-06,
+      "loss": 0.0899,
+      "step": 5046
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.340628582091543e-06,
+      "loss": 0.1104,
+      "step": 5047
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.340495294964414e-06,
+      "loss": 0.1016,
+      "step": 5048
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.340362007837283e-06,
+      "loss": 0.0987,
+      "step": 5049
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.340228720710156e-06,
+      "loss": 0.1366,
+      "step": 5050
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.340095433583025e-06,
+      "loss": 0.1069,
+      "step": 5051
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.339962146455896e-06,
+      "loss": 0.1257,
+      "step": 5052
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.339828859328767e-06,
+      "loss": 0.1393,
+      "step": 5053
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.339695572201638e-06,
+      "loss": 0.1147,
+      "step": 5054
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.339562285074509e-06,
+      "loss": 0.0934,
+      "step": 5055
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.33942899794738e-06,
+      "loss": 0.0748,
+      "step": 5056
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.33929571082025e-06,
+      "loss": 0.1435,
+      "step": 5057
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.33916242369312e-06,
+      "loss": 0.1001,
+      "step": 5058
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.339029136565991e-06,
+      "loss": 0.0989,
+      "step": 5059
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.338895849438862e-06,
+      "loss": 0.0975,
+      "step": 5060
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.338762562311733e-06,
+      "loss": 0.0919,
+      "step": 5061
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.338629275184603e-06,
+      "loss": 0.1622,
+      "step": 5062
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.338495988057474e-06,
+      "loss": 0.102,
+      "step": 5063
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.338362700930345e-06,
+      "loss": 0.0975,
+      "step": 5064
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.338229413803216e-06,
+      "loss": 0.1013,
+      "step": 5065
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.338096126676087e-06,
+      "loss": 0.132,
+      "step": 5066
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.337962839548958e-06,
+      "loss": 0.123,
+      "step": 5067
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.337829552421827e-06,
+      "loss": 0.0734,
+      "step": 5068
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.3376962652947e-06,
+      "loss": 0.0911,
+      "step": 5069
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.337562978167569e-06,
+      "loss": 0.1052,
+      "step": 5070
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.33742969104044e-06,
+      "loss": 0.1263,
+      "step": 5071
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.337296403913311e-06,
+      "loss": 0.1262,
+      "step": 5072
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.337163116786182e-06,
+      "loss": 0.1319,
+      "step": 5073
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.337029829659053e-06,
+      "loss": 0.0702,
+      "step": 5074
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.336896542531924e-06,
+      "loss": 0.1196,
+      "step": 5075
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.336763255404793e-06,
+      "loss": 0.1269,
+      "step": 5076
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.336629968277664e-06,
+      "loss": 0.0898,
+      "step": 5077
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.336496681150535e-06,
+      "loss": 0.1072,
+      "step": 5078
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.336363394023406e-06,
+      "loss": 0.0991,
+      "step": 5079
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.336230106896277e-06,
+      "loss": 0.1251,
+      "step": 5080
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.336096819769146e-06,
+      "loss": 0.1166,
+      "step": 5081
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.335963532642019e-06,
+      "loss": 0.0876,
+      "step": 5082
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.335830245514888e-06,
+      "loss": 0.1023,
+      "step": 5083
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.33569695838776e-06,
+      "loss": 0.0978,
+      "step": 5084
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.33556367126063e-06,
+      "loss": 0.1287,
+      "step": 5085
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.335430384133501e-06,
+      "loss": 0.1054,
+      "step": 5086
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.335297097006372e-06,
+      "loss": 0.1168,
+      "step": 5087
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.335163809879243e-06,
+      "loss": 0.109,
+      "step": 5088
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.335030522752113e-06,
+      "loss": 0.1137,
+      "step": 5089
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.334897235624984e-06,
+      "loss": 0.0884,
+      "step": 5090
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.334763948497855e-06,
+      "loss": 0.1049,
+      "step": 5091
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.334630661370726e-06,
+      "loss": 0.1442,
+      "step": 5092
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.334497374243597e-06,
+      "loss": 0.1328,
+      "step": 5093
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.334364087116468e-06,
+      "loss": 0.1223,
+      "step": 5094
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.334230799989337e-06,
+      "loss": 0.1144,
+      "step": 5095
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.334097512862208e-06,
+      "loss": 0.0989,
+      "step": 5096
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.333964225735079e-06,
+      "loss": 0.1487,
+      "step": 5097
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.33383093860795e-06,
+      "loss": 0.0961,
+      "step": 5098
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.333697651480821e-06,
+      "loss": 0.0868,
+      "step": 5099
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.33356436435369e-06,
+      "loss": 0.1094,
+      "step": 5100
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.333431077226563e-06,
+      "loss": 0.1209,
+      "step": 5101
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.333297790099432e-06,
+      "loss": 0.0771,
+      "step": 5102
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.333164502972303e-06,
+      "loss": 0.1225,
+      "step": 5103
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.333031215845174e-06,
+      "loss": 0.0997,
+      "step": 5104
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.332897928718045e-06,
+      "loss": 0.1648,
+      "step": 5105
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.332764641590916e-06,
+      "loss": 0.1077,
+      "step": 5106
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.332631354463787e-06,
+      "loss": 0.1056,
+      "step": 5107
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.332498067336657e-06,
+      "loss": 0.1385,
+      "step": 5108
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.33236478020953e-06,
+      "loss": 0.0981,
+      "step": 5109
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.332231493082399e-06,
+      "loss": 0.1297,
+      "step": 5110
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.33209820595527e-06,
+      "loss": 0.1184,
+      "step": 5111
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.33196491882814e-06,
+      "loss": 0.1046,
+      "step": 5112
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.331831631701011e-06,
+      "loss": 0.1108,
+      "step": 5113
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.331698344573882e-06,
+      "loss": 0.0764,
+      "step": 5114
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.331565057446752e-06,
+      "loss": 0.0992,
+      "step": 5115
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.331431770319623e-06,
+      "loss": 0.0941,
+      "step": 5116
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.331298483192494e-06,
+      "loss": 0.1329,
+      "step": 5117
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.331165196065365e-06,
+      "loss": 0.0899,
+      "step": 5118
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.331031908938236e-06,
+      "loss": 0.1458,
+      "step": 5119
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.330898621811107e-06,
+      "loss": 0.0868,
+      "step": 5120
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.330765334683976e-06,
+      "loss": 0.1203,
+      "step": 5121
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.330632047556849e-06,
+      "loss": 0.1172,
+      "step": 5122
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.330498760429718e-06,
+      "loss": 0.1461,
+      "step": 5123
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.330365473302589e-06,
+      "loss": 0.1092,
+      "step": 5124
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.33023218617546e-06,
+      "loss": 0.1079,
+      "step": 5125
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.330098899048331e-06,
+      "loss": 0.1022,
+      "step": 5126
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.329965611921202e-06,
+      "loss": 0.112,
+      "step": 5127
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.329832324794073e-06,
+      "loss": 0.0939,
+      "step": 5128
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.329699037666942e-06,
+      "loss": 0.1132,
+      "step": 5129
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.329565750539813e-06,
+      "loss": 0.0853,
+      "step": 5130
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.329432463412684e-06,
+      "loss": 0.0884,
+      "step": 5131
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.329299176285555e-06,
+      "loss": 0.0686,
+      "step": 5132
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.329165889158426e-06,
+      "loss": 0.1175,
+      "step": 5133
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.329032602031297e-06,
+      "loss": 0.1644,
+      "step": 5134
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.328899314904167e-06,
+      "loss": 0.074,
+      "step": 5135
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.328766027777038e-06,
+      "loss": 0.1481,
+      "step": 5136
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.328632740649909e-06,
+      "loss": 0.1316,
+      "step": 5137
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.32849945352278e-06,
+      "loss": 0.1279,
+      "step": 5138
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.32836616639565e-06,
+      "loss": 0.1382,
+      "step": 5139
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.32823287926852e-06,
+      "loss": 0.1091,
+      "step": 5140
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.328099592141393e-06,
+      "loss": 0.1231,
+      "step": 5141
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.327966305014262e-06,
+      "loss": 0.1023,
+      "step": 5142
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.327833017887133e-06,
+      "loss": 0.1137,
+      "step": 5143
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.327699730760004e-06,
+      "loss": 0.1356,
+      "step": 5144
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.327566443632875e-06,
+      "loss": 0.0758,
+      "step": 5145
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.327433156505746e-06,
+      "loss": 0.1122,
+      "step": 5146
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.327299869378617e-06,
+      "loss": 0.1391,
+      "step": 5147
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.327166582251486e-06,
+      "loss": 0.0722,
+      "step": 5148
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.327033295124359e-06,
+      "loss": 0.1198,
+      "step": 5149
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.326900007997228e-06,
+      "loss": 0.1054,
+      "step": 5150
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.326766720870099e-06,
+      "loss": 0.1375,
+      "step": 5151
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.32663343374297e-06,
+      "loss": 0.104,
+      "step": 5152
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.326500146615841e-06,
+      "loss": 0.1427,
+      "step": 5153
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.326366859488712e-06,
+      "loss": 0.1223,
+      "step": 5154
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.326233572361581e-06,
+      "loss": 0.1188,
+      "step": 5155
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.326100285234452e-06,
+      "loss": 0.1282,
+      "step": 5156
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.325966998107323e-06,
+      "loss": 0.0833,
+      "step": 5157
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.325833710980194e-06,
+      "loss": 0.1439,
+      "step": 5158
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.325700423853065e-06,
+      "loss": 0.0972,
+      "step": 5159
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.325567136725936e-06,
+      "loss": 0.1484,
+      "step": 5160
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.325433849598806e-06,
+      "loss": 0.117,
+      "step": 5161
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.325300562471677e-06,
+      "loss": 0.149,
+      "step": 5162
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.325167275344548e-06,
+      "loss": 0.1205,
+      "step": 5163
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.325033988217419e-06,
+      "loss": 0.1487,
+      "step": 5164
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.32490070109029e-06,
+      "loss": 0.1012,
+      "step": 5165
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.32476741396316e-06,
+      "loss": 0.1257,
+      "step": 5166
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.32463412683603e-06,
+      "loss": 0.1391,
+      "step": 5167
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.324500839708903e-06,
+      "loss": 0.1042,
+      "step": 5168
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.324367552581772e-06,
+      "loss": 0.1076,
+      "step": 5169
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.324234265454643e-06,
+      "loss": 0.1237,
+      "step": 5170
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.324100978327514e-06,
+      "loss": 0.1333,
+      "step": 5171
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.323967691200385e-06,
+      "loss": 0.105,
+      "step": 5172
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.323834404073256e-06,
+      "loss": 0.0824,
+      "step": 5173
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.323701116946125e-06,
+      "loss": 0.1414,
+      "step": 5174
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.323567829818996e-06,
+      "loss": 0.1068,
+      "step": 5175
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.323434542691867e-06,
+      "loss": 0.1203,
+      "step": 5176
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.323301255564738e-06,
+      "loss": 0.1228,
+      "step": 5177
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.32316796843761e-06,
+      "loss": 0.1249,
+      "step": 5178
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.32303468131048e-06,
+      "loss": 0.1225,
+      "step": 5179
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.32290139418335e-06,
+      "loss": 0.1546,
+      "step": 5180
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.322768107056222e-06,
+      "loss": 0.1058,
+      "step": 5181
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.322634819929091e-06,
+      "loss": 0.0918,
+      "step": 5182
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.322501532801962e-06,
+      "loss": 0.0913,
+      "step": 5183
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.322368245674833e-06,
+      "loss": 0.1363,
+      "step": 5184
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.322234958547704e-06,
+      "loss": 0.1475,
+      "step": 5185
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.322101671420575e-06,
+      "loss": 0.1081,
+      "step": 5186
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.321968384293446e-06,
+      "loss": 0.09,
+      "step": 5187
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.321835097166316e-06,
+      "loss": 0.1036,
+      "step": 5188
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.321701810039188e-06,
+      "loss": 0.1036,
+      "step": 5189
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.321568522912058e-06,
+      "loss": 0.097,
+      "step": 5190
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.321435235784929e-06,
+      "loss": 0.0971,
+      "step": 5191
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.3213019486578e-06,
+      "loss": 0.106,
+      "step": 5192
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.32116866153067e-06,
+      "loss": 0.1103,
+      "step": 5193
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.321035374403542e-06,
+      "loss": 0.0874,
+      "step": 5194
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.320902087276411e-06,
+      "loss": 0.1009,
+      "step": 5195
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.320768800149282e-06,
+      "loss": 0.0727,
+      "step": 5196
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.320635513022153e-06,
+      "loss": 0.1041,
+      "step": 5197
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.320502225895024e-06,
+      "loss": 0.1325,
+      "step": 5198
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.320368938767895e-06,
+      "loss": 0.091,
+      "step": 5199
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.320235651640766e-06,
+      "loss": 0.1282,
+      "step": 5200
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.320102364513635e-06,
+      "loss": 0.0773,
+      "step": 5201
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.319969077386506e-06,
+      "loss": 0.133,
+      "step": 5202
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.319835790259377e-06,
+      "loss": 0.1183,
+      "step": 5203
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.319702503132248e-06,
+      "loss": 0.1234,
+      "step": 5204
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.31956921600512e-06,
+      "loss": 0.1165,
+      "step": 5205
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.31943592887799e-06,
+      "loss": 0.0758,
+      "step": 5206
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.31930264175086e-06,
+      "loss": 0.1093,
+      "step": 5207
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.319169354623732e-06,
+      "loss": 0.1356,
+      "step": 5208
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.319036067496602e-06,
+      "loss": 0.1191,
+      "step": 5209
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.318902780369473e-06,
+      "loss": 0.1106,
+      "step": 5210
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.318769493242344e-06,
+      "loss": 0.1074,
+      "step": 5211
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.318636206115215e-06,
+      "loss": 0.1274,
+      "step": 5212
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.318502918988086e-06,
+      "loss": 0.1074,
+      "step": 5213
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.318369631860955e-06,
+      "loss": 0.1965,
+      "step": 5214
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.318236344733826e-06,
+      "loss": 0.1157,
+      "step": 5215
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.318103057606697e-06,
+      "loss": 0.0985,
+      "step": 5216
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.317969770479568e-06,
+      "loss": 0.1127,
+      "step": 5217
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.317836483352439e-06,
+      "loss": 0.1035,
+      "step": 5218
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.31770319622531e-06,
+      "loss": 0.1377,
+      "step": 5219
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.317569909098179e-06,
+      "loss": 0.1024,
+      "step": 5220
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.317436621971052e-06,
+      "loss": 0.0902,
+      "step": 5221
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.317303334843921e-06,
+      "loss": 0.1433,
+      "step": 5222
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.317170047716792e-06,
+      "loss": 0.1295,
+      "step": 5223
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.317036760589663e-06,
+      "loss": 0.1277,
+      "step": 5224
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.316903473462534e-06,
+      "loss": 0.1175,
+      "step": 5225
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.316770186335405e-06,
+      "loss": 0.1273,
+      "step": 5226
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.316636899208276e-06,
+      "loss": 0.1694,
+      "step": 5227
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.316503612081145e-06,
+      "loss": 0.1634,
+      "step": 5228
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.316370324954016e-06,
+      "loss": 0.1273,
+      "step": 5229
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.316237037826887e-06,
+      "loss": 0.1257,
+      "step": 5230
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.316103750699758e-06,
+      "loss": 0.0989,
+      "step": 5231
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.31597046357263e-06,
+      "loss": 0.1154,
+      "step": 5232
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.315837176445499e-06,
+      "loss": 0.1203,
+      "step": 5233
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.31570388931837e-06,
+      "loss": 0.1097,
+      "step": 5234
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.31557060219124e-06,
+      "loss": 0.1072,
+      "step": 5235
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.315437315064112e-06,
+      "loss": 0.1349,
+      "step": 5236
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.315304027936983e-06,
+      "loss": 0.0858,
+      "step": 5237
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.315170740809854e-06,
+      "loss": 0.1212,
+      "step": 5238
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.315037453682723e-06,
+      "loss": 0.1462,
+      "step": 5239
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.314904166555596e-06,
+      "loss": 0.1252,
+      "step": 5240
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.314770879428465e-06,
+      "loss": 0.1032,
+      "step": 5241
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.314637592301336e-06,
+      "loss": 0.1289,
+      "step": 5242
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.314504305174207e-06,
+      "loss": 0.1084,
+      "step": 5243
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.314371018047078e-06,
+      "loss": 0.0998,
+      "step": 5244
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.314237730919949e-06,
+      "loss": 0.1087,
+      "step": 5245
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.31410444379282e-06,
+      "loss": 0.1248,
+      "step": 5246
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.31397115666569e-06,
+      "loss": 0.0901,
+      "step": 5247
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.313837869538562e-06,
+      "loss": 0.1125,
+      "step": 5248
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.313704582411431e-06,
+      "loss": 0.074,
+      "step": 5249
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.313571295284302e-06,
+      "loss": 0.1473,
+      "step": 5250
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.313438008157173e-06,
+      "loss": 0.0848,
+      "step": 5251
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.313304721030042e-06,
+      "loss": 0.1641,
+      "step": 5252
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.313171433902915e-06,
+      "loss": 0.0733,
+      "step": 5253
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.313038146775784e-06,
+      "loss": 0.1829,
+      "step": 5254
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.312904859648655e-06,
+      "loss": 0.1184,
+      "step": 5255
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.312771572521526e-06,
+      "loss": 0.2204,
+      "step": 5256
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.312638285394397e-06,
+      "loss": 0.1542,
+      "step": 5257
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.312504998267268e-06,
+      "loss": 0.1417,
+      "step": 5258
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.31237171114014e-06,
+      "loss": 0.1349,
+      "step": 5259
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.312238424013009e-06,
+      "loss": 0.1003,
+      "step": 5260
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.31210513688588e-06,
+      "loss": 0.1267,
+      "step": 5261
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.31197184975875e-06,
+      "loss": 0.1123,
+      "step": 5262
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.311838562631622e-06,
+      "loss": 0.1171,
+      "step": 5263
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.311705275504493e-06,
+      "loss": 0.107,
+      "step": 5264
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.311571988377364e-06,
+      "loss": 0.1388,
+      "step": 5265
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.311438701250233e-06,
+      "loss": 0.1024,
+      "step": 5266
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.311305414123106e-06,
+      "loss": 0.12,
+      "step": 5267
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.311172126995975e-06,
+      "loss": 0.1124,
+      "step": 5268
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.311038839868846e-06,
+      "loss": 0.1098,
+      "step": 5269
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.310905552741717e-06,
+      "loss": 0.1,
+      "step": 5270
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.310772265614588e-06,
+      "loss": 0.0893,
+      "step": 5271
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.310638978487459e-06,
+      "loss": 0.1283,
+      "step": 5272
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.310505691360328e-06,
+      "loss": 0.1265,
+      "step": 5273
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.3103724042332e-06,
+      "loss": 0.1182,
+      "step": 5274
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.31023911710607e-06,
+      "loss": 0.1086,
+      "step": 5275
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.310105829978941e-06,
+      "loss": 0.1015,
+      "step": 5276
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.309972542851812e-06,
+      "loss": 0.0857,
+      "step": 5277
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.309839255724683e-06,
+      "loss": 0.1046,
+      "step": 5278
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.309705968597553e-06,
+      "loss": 0.1188,
+      "step": 5279
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.309572681470425e-06,
+      "loss": 0.0995,
+      "step": 5280
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.309439394343295e-06,
+      "loss": 0.0979,
+      "step": 5281
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.309306107216166e-06,
+      "loss": 0.084,
+      "step": 5282
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.309172820089037e-06,
+      "loss": 0.0922,
+      "step": 5283
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.309039532961908e-06,
+      "loss": 0.081,
+      "step": 5284
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.308906245834779e-06,
+      "loss": 0.1049,
+      "step": 5285
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.30877295870765e-06,
+      "loss": 0.1426,
+      "step": 5286
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.308639671580519e-06,
+      "loss": 0.1027,
+      "step": 5287
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.308506384453392e-06,
+      "loss": 0.1449,
+      "step": 5288
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.30837309732626e-06,
+      "loss": 0.1017,
+      "step": 5289
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.308239810199132e-06,
+      "loss": 0.1205,
+      "step": 5290
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.308106523072003e-06,
+      "loss": 0.0758,
+      "step": 5291
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.307973235944872e-06,
+      "loss": 0.079,
+      "step": 5292
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.307839948817745e-06,
+      "loss": 0.0966,
+      "step": 5293
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.307706661690614e-06,
+      "loss": 0.1039,
+      "step": 5294
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.307573374563485e-06,
+      "loss": 0.1221,
+      "step": 5295
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.307440087436356e-06,
+      "loss": 0.0984,
+      "step": 5296
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.307306800309227e-06,
+      "loss": 0.1015,
+      "step": 5297
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.307173513182098e-06,
+      "loss": 0.0788,
+      "step": 5298
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.307040226054969e-06,
+      "loss": 0.0962,
+      "step": 5299
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.306906938927838e-06,
+      "loss": 0.098,
+      "step": 5300
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.30677365180071e-06,
+      "loss": 0.1352,
+      "step": 5301
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.30664036467358e-06,
+      "loss": 0.1421,
+      "step": 5302
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.306507077546451e-06,
+      "loss": 0.1107,
+      "step": 5303
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.306373790419322e-06,
+      "loss": 0.1362,
+      "step": 5304
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.306240503292193e-06,
+      "loss": 0.113,
+      "step": 5305
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.306107216165063e-06,
+      "loss": 0.1105,
+      "step": 5306
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.305973929037935e-06,
+      "loss": 0.1231,
+      "step": 5307
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.305840641910805e-06,
+      "loss": 0.1451,
+      "step": 5308
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.305707354783676e-06,
+      "loss": 0.1073,
+      "step": 5309
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.305574067656547e-06,
+      "loss": 0.1162,
+      "step": 5310
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.305440780529416e-06,
+      "loss": 0.1117,
+      "step": 5311
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.305307493402289e-06,
+      "loss": 0.1378,
+      "step": 5312
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.305174206275158e-06,
+      "loss": 0.1369,
+      "step": 5313
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.305040919148029e-06,
+      "loss": 0.1259,
+      "step": 5314
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.3049076320209e-06,
+      "loss": 0.0766,
+      "step": 5315
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.304774344893771e-06,
+      "loss": 0.1422,
+      "step": 5316
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.304641057766642e-06,
+      "loss": 0.1555,
+      "step": 5317
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.304507770639513e-06,
+      "loss": 0.0998,
+      "step": 5318
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.304374483512382e-06,
+      "loss": 0.0762,
+      "step": 5319
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.304241196385255e-06,
+      "loss": 0.0861,
+      "step": 5320
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.304107909258124e-06,
+      "loss": 0.1524,
+      "step": 5321
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.303974622130995e-06,
+      "loss": 0.1141,
+      "step": 5322
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.303841335003866e-06,
+      "loss": 0.1009,
+      "step": 5323
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.303708047876737e-06,
+      "loss": 0.101,
+      "step": 5324
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.303574760749608e-06,
+      "loss": 0.0878,
+      "step": 5325
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.303441473622479e-06,
+      "loss": 0.128,
+      "step": 5326
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.303308186495348e-06,
+      "loss": 0.107,
+      "step": 5327
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.30317489936822e-06,
+      "loss": 0.0967,
+      "step": 5328
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.30304161224109e-06,
+      "loss": 0.0953,
+      "step": 5329
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.302908325113961e-06,
+      "loss": 0.0862,
+      "step": 5330
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.302775037986832e-06,
+      "loss": 0.1502,
+      "step": 5331
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.302641750859702e-06,
+      "loss": 0.1387,
+      "step": 5332
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.302508463732573e-06,
+      "loss": 0.1028,
+      "step": 5333
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.302375176605444e-06,
+      "loss": 0.1328,
+      "step": 5334
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.302241889478315e-06,
+      "loss": 0.1246,
+      "step": 5335
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.302108602351186e-06,
+      "loss": 0.1144,
+      "step": 5336
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.301975315224057e-06,
+      "loss": 0.0929,
+      "step": 5337
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.301842028096926e-06,
+      "loss": 0.1675,
+      "step": 5338
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.301708740969799e-06,
+      "loss": 0.131,
+      "step": 5339
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.301575453842668e-06,
+      "loss": 0.1102,
+      "step": 5340
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.301442166715539e-06,
+      "loss": 0.1249,
+      "step": 5341
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.30130887958841e-06,
+      "loss": 0.1227,
+      "step": 5342
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.301175592461281e-06,
+      "loss": 0.0875,
+      "step": 5343
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.301042305334152e-06,
+      "loss": 0.0938,
+      "step": 5344
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.300909018207023e-06,
+      "loss": 0.1217,
+      "step": 5345
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.300775731079892e-06,
+      "loss": 0.1516,
+      "step": 5346
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.300642443952765e-06,
+      "loss": 0.1831,
+      "step": 5347
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.300509156825634e-06,
+      "loss": 0.15,
+      "step": 5348
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.300375869698505e-06,
+      "loss": 0.113,
+      "step": 5349
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.300242582571376e-06,
+      "loss": 0.0778,
+      "step": 5350
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.300109295444246e-06,
+      "loss": 0.1005,
+      "step": 5351
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.299976008317118e-06,
+      "loss": 0.0897,
+      "step": 5352
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.299842721189988e-06,
+      "loss": 0.1161,
+      "step": 5353
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.299709434062859e-06,
+      "loss": 0.1298,
+      "step": 5354
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.29957614693573e-06,
+      "loss": 0.1196,
+      "step": 5355
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.2994428598086e-06,
+      "loss": 0.1003,
+      "step": 5356
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.299309572681472e-06,
+      "loss": 0.1146,
+      "step": 5357
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.299176285554343e-06,
+      "loss": 0.0878,
+      "step": 5358
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.299042998427212e-06,
+      "loss": 0.1075,
+      "step": 5359
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.298909711300084e-06,
+      "loss": 0.1379,
+      "step": 5360
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.298776424172954e-06,
+      "loss": 0.1657,
+      "step": 5361
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.298643137045825e-06,
+      "loss": 0.1049,
+      "step": 5362
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.298509849918696e-06,
+      "loss": 0.1086,
+      "step": 5363
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.298376562791567e-06,
+      "loss": 0.1199,
+      "step": 5364
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.298243275664438e-06,
+      "loss": 0.1043,
+      "step": 5365
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.298109988537309e-06,
+      "loss": 0.1019,
+      "step": 5366
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.297976701410178e-06,
+      "loss": 0.1115,
+      "step": 5367
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.297843414283049e-06,
+      "loss": 0.1156,
+      "step": 5368
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.29771012715592e-06,
+      "loss": 0.0871,
+      "step": 5369
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.297576840028791e-06,
+      "loss": 0.0916,
+      "step": 5370
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.297443552901662e-06,
+      "loss": 0.1079,
+      "step": 5371
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.297310265774531e-06,
+      "loss": 0.126,
+      "step": 5372
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.297176978647402e-06,
+      "loss": 0.1554,
+      "step": 5373
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.297043691520273e-06,
+      "loss": 0.1216,
+      "step": 5374
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.296910404393144e-06,
+      "loss": 0.1044,
+      "step": 5375
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.296777117266015e-06,
+      "loss": 0.1521,
+      "step": 5376
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.296643830138886e-06,
+      "loss": 0.1059,
+      "step": 5377
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.296510543011756e-06,
+      "loss": 0.0924,
+      "step": 5378
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.296377255884628e-06,
+      "loss": 0.0881,
+      "step": 5379
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.296243968757498e-06,
+      "loss": 0.0989,
+      "step": 5380
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.296110681630369e-06,
+      "loss": 0.1382,
+      "step": 5381
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.29597739450324e-06,
+      "loss": 0.0998,
+      "step": 5382
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.29584410737611e-06,
+      "loss": 0.1231,
+      "step": 5383
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.295710820248982e-06,
+      "loss": 0.0802,
+      "step": 5384
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.295577533121853e-06,
+      "loss": 0.1045,
+      "step": 5385
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.295444245994722e-06,
+      "loss": 0.1859,
+      "step": 5386
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.295310958867595e-06,
+      "loss": 0.0958,
+      "step": 5387
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.295177671740464e-06,
+      "loss": 0.1063,
+      "step": 5388
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.295044384613335e-06,
+      "loss": 0.1203,
+      "step": 5389
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.294911097486206e-06,
+      "loss": 0.112,
+      "step": 5390
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.294777810359075e-06,
+      "loss": 0.1031,
+      "step": 5391
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.294644523231948e-06,
+      "loss": 0.1411,
+      "step": 5392
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.294511236104817e-06,
+      "loss": 0.0849,
+      "step": 5393
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.294377948977688e-06,
+      "loss": 0.1411,
+      "step": 5394
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.294244661850559e-06,
+      "loss": 0.0945,
+      "step": 5395
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.29411137472343e-06,
+      "loss": 0.0817,
+      "step": 5396
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.293978087596301e-06,
+      "loss": 0.086,
+      "step": 5397
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.293844800469172e-06,
+      "loss": 0.1543,
+      "step": 5398
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.293711513342041e-06,
+      "loss": 0.1196,
+      "step": 5399
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.293578226214912e-06,
+      "loss": 0.1264,
+      "step": 5400
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.293444939087783e-06,
+      "loss": 0.1254,
+      "step": 5401
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.293311651960654e-06,
+      "loss": 0.0923,
+      "step": 5402
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.293178364833525e-06,
+      "loss": 0.0998,
+      "step": 5403
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.293045077706396e-06,
+      "loss": 0.1503,
+      "step": 5404
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.292911790579266e-06,
+      "loss": 0.0875,
+      "step": 5405
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.292778503452138e-06,
+      "loss": 0.0875,
+      "step": 5406
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.292645216325008e-06,
+      "loss": 0.1281,
+      "step": 5407
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.292511929197879e-06,
+      "loss": 0.1239,
+      "step": 5408
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.29237864207075e-06,
+      "loss": 0.1236,
+      "step": 5409
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.292245354943619e-06,
+      "loss": 0.0979,
+      "step": 5410
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.292112067816492e-06,
+      "loss": 0.1426,
+      "step": 5411
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.291978780689361e-06,
+      "loss": 0.111,
+      "step": 5412
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.291845493562232e-06,
+      "loss": 0.1487,
+      "step": 5413
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.291712206435103e-06,
+      "loss": 0.0848,
+      "step": 5414
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.291578919307974e-06,
+      "loss": 0.1228,
+      "step": 5415
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.291445632180845e-06,
+      "loss": 0.0923,
+      "step": 5416
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.291312345053716e-06,
+      "loss": 0.163,
+      "step": 5417
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.291179057926585e-06,
+      "loss": 0.1037,
+      "step": 5418
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.291045770799458e-06,
+      "loss": 0.091,
+      "step": 5419
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.290912483672327e-06,
+      "loss": 0.0997,
+      "step": 5420
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.290779196545198e-06,
+      "loss": 0.1173,
+      "step": 5421
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.29064590941807e-06,
+      "loss": 0.1291,
+      "step": 5422
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.29051262229094e-06,
+      "loss": 0.084,
+      "step": 5423
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.290379335163811e-06,
+      "loss": 0.0905,
+      "step": 5424
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.290246048036682e-06,
+      "loss": 0.1156,
+      "step": 5425
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.290112760909552e-06,
+      "loss": 0.1479,
+      "step": 5426
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.289979473782424e-06,
+      "loss": 0.1047,
+      "step": 5427
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.289846186655294e-06,
+      "loss": 0.0822,
+      "step": 5428
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.289712899528164e-06,
+      "loss": 0.0804,
+      "step": 5429
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.289579612401035e-06,
+      "loss": 0.1167,
+      "step": 5430
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.289446325273905e-06,
+      "loss": 0.124,
+      "step": 5431
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.289313038146777e-06,
+      "loss": 0.1133,
+      "step": 5432
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.289179751019647e-06,
+      "loss": 0.1087,
+      "step": 5433
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.289046463892518e-06,
+      "loss": 0.0906,
+      "step": 5434
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.288913176765389e-06,
+      "loss": 0.1266,
+      "step": 5435
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.28877988963826e-06,
+      "loss": 0.106,
+      "step": 5436
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.28864660251113e-06,
+      "loss": 0.0825,
+      "step": 5437
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.288513315384002e-06,
+      "loss": 0.1275,
+      "step": 5438
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.288380028256871e-06,
+      "loss": 0.1666,
+      "step": 5439
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.288246741129742e-06,
+      "loss": 0.095,
+      "step": 5440
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.288113454002613e-06,
+      "loss": 0.0894,
+      "step": 5441
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.287980166875484e-06,
+      "loss": 0.1197,
+      "step": 5442
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.287846879748355e-06,
+      "loss": 0.1115,
+      "step": 5443
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.287713592621226e-06,
+      "loss": 0.0983,
+      "step": 5444
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.287580305494095e-06,
+      "loss": 0.1283,
+      "step": 5445
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.287447018366968e-06,
+      "loss": 0.0999,
+      "step": 5446
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.287313731239837e-06,
+      "loss": 0.1098,
+      "step": 5447
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.287180444112708e-06,
+      "loss": 0.1133,
+      "step": 5448
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.28704715698558e-06,
+      "loss": 0.1093,
+      "step": 5449
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.286913869858449e-06,
+      "loss": 0.1453,
+      "step": 5450
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.286780582731321e-06,
+      "loss": 0.0864,
+      "step": 5451
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.28664729560419e-06,
+      "loss": 0.0909,
+      "step": 5452
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.286514008477062e-06,
+      "loss": 0.0826,
+      "step": 5453
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.286380721349933e-06,
+      "loss": 0.158,
+      "step": 5454
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.286247434222804e-06,
+      "loss": 0.0991,
+      "step": 5455
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.286114147095675e-06,
+      "loss": 0.1262,
+      "step": 5456
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.285980859968546e-06,
+      "loss": 0.1036,
+      "step": 5457
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.285847572841415e-06,
+      "loss": 0.0858,
+      "step": 5458
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.285714285714288e-06,
+      "loss": 0.1193,
+      "step": 5459
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.285580998587157e-06,
+      "loss": 0.0958,
+      "step": 5460
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.285447711460028e-06,
+      "loss": 0.1403,
+      "step": 5461
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.285314424332899e-06,
+      "loss": 0.0819,
+      "step": 5462
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.28518113720577e-06,
+      "loss": 0.1274,
+      "step": 5463
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.28504785007864e-06,
+      "loss": 0.1174,
+      "step": 5464
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.284914562951512e-06,
+      "loss": 0.1233,
+      "step": 5465
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.284781275824381e-06,
+      "loss": 0.0783,
+      "step": 5466
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.284647988697252e-06,
+      "loss": 0.0633,
+      "step": 5467
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.284514701570123e-06,
+      "loss": 0.1143,
+      "step": 5468
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.284381414442994e-06,
+      "loss": 0.1033,
+      "step": 5469
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.284248127315865e-06,
+      "loss": 0.1184,
+      "step": 5470
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.284114840188734e-06,
+      "loss": 0.0981,
+      "step": 5471
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.283981553061605e-06,
+      "loss": 0.1058,
+      "step": 5472
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.283848265934476e-06,
+      "loss": 0.0868,
+      "step": 5473
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.283714978807347e-06,
+      "loss": 0.1099,
+      "step": 5474
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.283581691680218e-06,
+      "loss": 0.1092,
+      "step": 5475
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.28344840455309e-06,
+      "loss": 0.1356,
+      "step": 5476
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.283315117425959e-06,
+      "loss": 0.1262,
+      "step": 5477
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.283181830298831e-06,
+      "loss": 0.1197,
+      "step": 5478
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.2830485431717e-06,
+      "loss": 0.1016,
+      "step": 5479
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.282915256044572e-06,
+      "loss": 0.1168,
+      "step": 5480
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.282781968917443e-06,
+      "loss": 0.102,
+      "step": 5481
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.282648681790314e-06,
+      "loss": 0.1014,
+      "step": 5482
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.282515394663185e-06,
+      "loss": 0.1123,
+      "step": 5483
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.282382107536056e-06,
+      "loss": 0.108,
+      "step": 5484
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.282248820408925e-06,
+      "loss": 0.1239,
+      "step": 5485
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.282115533281798e-06,
+      "loss": 0.0947,
+      "step": 5486
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.281982246154667e-06,
+      "loss": 0.1189,
+      "step": 5487
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.281848959027538e-06,
+      "loss": 0.1495,
+      "step": 5488
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.281715671900409e-06,
+      "loss": 0.1192,
+      "step": 5489
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.281582384773278e-06,
+      "loss": 0.1136,
+      "step": 5490
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.281449097646151e-06,
+      "loss": 0.0815,
+      "step": 5491
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.28131581051902e-06,
+      "loss": 0.0939,
+      "step": 5492
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.281182523391891e-06,
+      "loss": 0.0981,
+      "step": 5493
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.281049236264762e-06,
+      "loss": 0.1385,
+      "step": 5494
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.280915949137633e-06,
+      "loss": 0.0746,
+      "step": 5495
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.280782662010504e-06,
+      "loss": 0.0984,
+      "step": 5496
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.280649374883375e-06,
+      "loss": 0.1157,
+      "step": 5497
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.280516087756244e-06,
+      "loss": 0.0979,
+      "step": 5498
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.280382800629115e-06,
+      "loss": 0.097,
+      "step": 5499
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.280249513501986e-06,
+      "loss": 0.096,
+      "step": 5500
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.280116226374857e-06,
+      "loss": 0.0754,
+      "step": 5501
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.279982939247728e-06,
+      "loss": 0.103,
+      "step": 5502
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.2798496521206e-06,
+      "loss": 0.1133,
+      "step": 5503
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.279716364993469e-06,
+      "loss": 0.1351,
+      "step": 5504
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.279583077866341e-06,
+      "loss": 0.1298,
+      "step": 5505
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.27944979073921e-06,
+      "loss": 0.0887,
+      "step": 5506
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.279316503612082e-06,
+      "loss": 0.1069,
+      "step": 5507
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.279183216484953e-06,
+      "loss": 0.1081,
+      "step": 5508
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.279049929357822e-06,
+      "loss": 0.1241,
+      "step": 5509
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.278916642230695e-06,
+      "loss": 0.0963,
+      "step": 5510
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.278783355103564e-06,
+      "loss": 0.1403,
+      "step": 5511
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.278650067976435e-06,
+      "loss": 0.1072,
+      "step": 5512
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.278516780849306e-06,
+      "loss": 0.117,
+      "step": 5513
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.278383493722177e-06,
+      "loss": 0.1021,
+      "step": 5514
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.278250206595048e-06,
+      "loss": 0.1041,
+      "step": 5515
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.278116919467919e-06,
+      "loss": 0.1134,
+      "step": 5516
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.277983632340788e-06,
+      "loss": 0.1029,
+      "step": 5517
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.277850345213661e-06,
+      "loss": 0.1526,
+      "step": 5518
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.27771705808653e-06,
+      "loss": 0.0935,
+      "step": 5519
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.277583770959401e-06,
+      "loss": 0.1166,
+      "step": 5520
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.277450483832272e-06,
+      "loss": 0.0974,
+      "step": 5521
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.277317196705143e-06,
+      "loss": 0.1032,
+      "step": 5522
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.277183909578014e-06,
+      "loss": 0.1077,
+      "step": 5523
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.277050622450885e-06,
+      "loss": 0.0762,
+      "step": 5524
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.276917335323755e-06,
+      "loss": 0.1317,
+      "step": 5525
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.276784048196627e-06,
+      "loss": 0.0869,
+      "step": 5526
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.276650761069497e-06,
+      "loss": 0.0873,
+      "step": 5527
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.276517473942368e-06,
+      "loss": 0.1137,
+      "step": 5528
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.276384186815239e-06,
+      "loss": 0.0961,
+      "step": 5529
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.276250899688108e-06,
+      "loss": 0.1319,
+      "step": 5530
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.27611761256098e-06,
+      "loss": 0.1045,
+      "step": 5531
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.27598432543385e-06,
+      "loss": 0.0813,
+      "step": 5532
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.27585103830672e-06,
+      "loss": 0.1193,
+      "step": 5533
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.275717751179592e-06,
+      "loss": 0.1139,
+      "step": 5534
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.275584464052463e-06,
+      "loss": 0.1238,
+      "step": 5535
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.275451176925334e-06,
+      "loss": 0.1175,
+      "step": 5536
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.275317889798205e-06,
+      "loss": 0.1038,
+      "step": 5537
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.275184602671074e-06,
+      "loss": 0.0991,
+      "step": 5538
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.275051315543945e-06,
+      "loss": 0.1172,
+      "step": 5539
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.274918028416816e-06,
+      "loss": 0.1367,
+      "step": 5540
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.274784741289687e-06,
+      "loss": 0.1647,
+      "step": 5541
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.274651454162558e-06,
+      "loss": 0.1176,
+      "step": 5542
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.274518167035429e-06,
+      "loss": 0.0978,
+      "step": 5543
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.274384879908298e-06,
+      "loss": 0.124,
+      "step": 5544
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.274251592781171e-06,
+      "loss": 0.1001,
+      "step": 5545
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.27411830565404e-06,
+      "loss": 0.1039,
+      "step": 5546
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.273985018526911e-06,
+      "loss": 0.1051,
+      "step": 5547
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.273851731399782e-06,
+      "loss": 0.073,
+      "step": 5548
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.273718444272652e-06,
+      "loss": 0.0687,
+      "step": 5549
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.273585157145524e-06,
+      "loss": 0.113,
+      "step": 5550
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.273451870018394e-06,
+      "loss": 0.0649,
+      "step": 5551
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.273318582891265e-06,
+      "loss": 0.0866,
+      "step": 5552
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.273185295764136e-06,
+      "loss": 0.1137,
+      "step": 5553
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.273052008637007e-06,
+      "loss": 0.1302,
+      "step": 5554
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.272918721509878e-06,
+      "loss": 0.1533,
+      "step": 5555
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.272785434382749e-06,
+      "loss": 0.117,
+      "step": 5556
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.272652147255618e-06,
+      "loss": 0.0947,
+      "step": 5557
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.27251886012849e-06,
+      "loss": 0.1051,
+      "step": 5558
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.27238557300136e-06,
+      "loss": 0.1247,
+      "step": 5559
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.272252285874231e-06,
+      "loss": 0.099,
+      "step": 5560
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.272118998747102e-06,
+      "loss": 0.1332,
+      "step": 5561
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.271985711619973e-06,
+      "loss": 0.1007,
+      "step": 5562
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.271852424492844e-06,
+      "loss": 0.0819,
+      "step": 5563
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.271719137365715e-06,
+      "loss": 0.0907,
+      "step": 5564
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.271585850238584e-06,
+      "loss": 0.0598,
+      "step": 5565
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.271452563111455e-06,
+      "loss": 0.1303,
+      "step": 5566
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.271319275984326e-06,
+      "loss": 0.1484,
+      "step": 5567
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.271185988857197e-06,
+      "loss": 0.1006,
+      "step": 5568
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.271052701730068e-06,
+      "loss": 0.1154,
+      "step": 5569
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.270919414602937e-06,
+      "loss": 0.0935,
+      "step": 5570
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.270786127475808e-06,
+      "loss": 0.1282,
+      "step": 5571
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.27065284034868e-06,
+      "loss": 0.0781,
+      "step": 5572
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.27051955322155e-06,
+      "loss": 0.0953,
+      "step": 5573
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.270386266094421e-06,
+      "loss": 0.1062,
+      "step": 5574
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.270252978967292e-06,
+      "loss": 0.0788,
+      "step": 5575
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.270119691840162e-06,
+      "loss": 0.1012,
+      "step": 5576
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.269986404713034e-06,
+      "loss": 0.1299,
+      "step": 5577
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.269853117585904e-06,
+      "loss": 0.1373,
+      "step": 5578
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.269719830458775e-06,
+      "loss": 0.0964,
+      "step": 5579
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.269586543331646e-06,
+      "loss": 0.0946,
+      "step": 5580
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.269453256204517e-06,
+      "loss": 0.1293,
+      "step": 5581
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.269319969077388e-06,
+      "loss": 0.1184,
+      "step": 5582
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.269186681950259e-06,
+      "loss": 0.1161,
+      "step": 5583
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.269053394823128e-06,
+      "loss": 0.1479,
+      "step": 5584
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.268920107695999e-06,
+      "loss": 0.0937,
+      "step": 5585
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.26878682056887e-06,
+      "loss": 0.1177,
+      "step": 5586
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.268653533441741e-06,
+      "loss": 0.1166,
+      "step": 5587
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.268520246314612e-06,
+      "loss": 0.0903,
+      "step": 5588
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.268386959187481e-06,
+      "loss": 0.0955,
+      "step": 5589
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.268253672060354e-06,
+      "loss": 0.1016,
+      "step": 5590
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.268120384933223e-06,
+      "loss": 0.0934,
+      "step": 5591
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.267987097806094e-06,
+      "loss": 0.0893,
+      "step": 5592
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.267853810678965e-06,
+      "loss": 0.1122,
+      "step": 5593
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.267720523551836e-06,
+      "loss": 0.1117,
+      "step": 5594
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.267587236424707e-06,
+      "loss": 0.0965,
+      "step": 5595
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.267453949297578e-06,
+      "loss": 0.1288,
+      "step": 5596
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.267320662170448e-06,
+      "loss": 0.1026,
+      "step": 5597
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.26718737504332e-06,
+      "loss": 0.0852,
+      "step": 5598
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.26705408791619e-06,
+      "loss": 0.0993,
+      "step": 5599
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.26692080078906e-06,
+      "loss": 0.1094,
+      "step": 5600
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.266787513661932e-06,
+      "loss": 0.1063,
+      "step": 5601
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.266654226534803e-06,
+      "loss": 0.121,
+      "step": 5602
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.266520939407674e-06,
+      "loss": 0.0856,
+      "step": 5603
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.266387652280545e-06,
+      "loss": 0.0979,
+      "step": 5604
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.266254365153414e-06,
+      "loss": 0.1356,
+      "step": 5605
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.266121078026285e-06,
+      "loss": 0.1102,
+      "step": 5606
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.265987790899156e-06,
+      "loss": 0.0729,
+      "step": 5607
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.265854503772027e-06,
+      "loss": 0.1111,
+      "step": 5608
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.265721216644898e-06,
+      "loss": 0.1296,
+      "step": 5609
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.265587929517767e-06,
+      "loss": 0.1279,
+      "step": 5610
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.265454642390638e-06,
+      "loss": 0.1052,
+      "step": 5611
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.265321355263509e-06,
+      "loss": 0.1348,
+      "step": 5612
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.26518806813638e-06,
+      "loss": 0.1241,
+      "step": 5613
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.265054781009251e-06,
+      "loss": 0.1203,
+      "step": 5614
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.264921493882122e-06,
+      "loss": 0.1029,
+      "step": 5615
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.264788206754991e-06,
+      "loss": 0.0818,
+      "step": 5616
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.264654919627864e-06,
+      "loss": 0.1043,
+      "step": 5617
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.264521632500733e-06,
+      "loss": 0.0894,
+      "step": 5618
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.264388345373604e-06,
+      "loss": 0.1609,
+      "step": 5619
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.264255058246475e-06,
+      "loss": 0.0869,
+      "step": 5620
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.264121771119346e-06,
+      "loss": 0.1236,
+      "step": 5621
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.263988483992217e-06,
+      "loss": 0.1573,
+      "step": 5622
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.263855196865088e-06,
+      "loss": 0.097,
+      "step": 5623
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.263721909737958e-06,
+      "loss": 0.1626,
+      "step": 5624
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.263588622610829e-06,
+      "loss": 0.104,
+      "step": 5625
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.2634553354837e-06,
+      "loss": 0.094,
+      "step": 5626
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.26332204835657e-06,
+      "loss": 0.0926,
+      "step": 5627
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.263188761229442e-06,
+      "loss": 0.1187,
+      "step": 5628
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.263055474102311e-06,
+      "loss": 0.117,
+      "step": 5629
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.262922186975184e-06,
+      "loss": 0.141,
+      "step": 5630
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.262788899848053e-06,
+      "loss": 0.0906,
+      "step": 5631
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.262655612720924e-06,
+      "loss": 0.1083,
+      "step": 5632
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.262522325593795e-06,
+      "loss": 0.1154,
+      "step": 5633
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.262389038466666e-06,
+      "loss": 0.122,
+      "step": 5634
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.262255751339537e-06,
+      "loss": 0.1477,
+      "step": 5635
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.262122464212408e-06,
+      "loss": 0.1159,
+      "step": 5636
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.261989177085277e-06,
+      "loss": 0.1052,
+      "step": 5637
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.261855889958148e-06,
+      "loss": 0.1252,
+      "step": 5638
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.26172260283102e-06,
+      "loss": 0.1058,
+      "step": 5639
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.26158931570389e-06,
+      "loss": 0.1206,
+      "step": 5640
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.261456028576761e-06,
+      "loss": 0.0673,
+      "step": 5641
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.261322741449632e-06,
+      "loss": 0.0812,
+      "step": 5642
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.261189454322501e-06,
+      "loss": 0.0837,
+      "step": 5643
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.261056167195372e-06,
+      "loss": 0.0999,
+      "step": 5644
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.260922880068243e-06,
+      "loss": 0.1107,
+      "step": 5645
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.260789592941114e-06,
+      "loss": 0.1187,
+      "step": 5646
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.260656305813985e-06,
+      "loss": 0.1181,
+      "step": 5647
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.260523018686855e-06,
+      "loss": 0.091,
+      "step": 5648
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.260389731559727e-06,
+      "loss": 0.0981,
+      "step": 5649
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.260256444432597e-06,
+      "loss": 0.1175,
+      "step": 5650
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.260123157305468e-06,
+      "loss": 0.0886,
+      "step": 5651
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.259989870178339e-06,
+      "loss": 0.1357,
+      "step": 5652
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.25985658305121e-06,
+      "loss": 0.0869,
+      "step": 5653
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.25972329592408e-06,
+      "loss": 0.0925,
+      "step": 5654
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.259590008796952e-06,
+      "loss": 0.1154,
+      "step": 5655
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.259456721669821e-06,
+      "loss": 0.1141,
+      "step": 5656
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.259323434542694e-06,
+      "loss": 0.0806,
+      "step": 5657
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.259190147415563e-06,
+      "loss": 0.1053,
+      "step": 5658
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.259056860288434e-06,
+      "loss": 0.1226,
+      "step": 5659
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.258923573161305e-06,
+      "loss": 0.1118,
+      "step": 5660
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.258790286034176e-06,
+      "loss": 0.1301,
+      "step": 5661
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.258656998907047e-06,
+      "loss": 0.0879,
+      "step": 5662
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.258523711779918e-06,
+      "loss": 0.1123,
+      "step": 5663
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.258390424652787e-06,
+      "loss": 0.0994,
+      "step": 5664
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.258257137525658e-06,
+      "loss": 0.1228,
+      "step": 5665
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.25812385039853e-06,
+      "loss": 0.1348,
+      "step": 5666
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.2579905632714e-06,
+      "loss": 0.0993,
+      "step": 5667
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.257857276144271e-06,
+      "loss": 0.1229,
+      "step": 5668
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.25772398901714e-06,
+      "loss": 0.1088,
+      "step": 5669
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.257590701890013e-06,
+      "loss": 0.1331,
+      "step": 5670
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.257457414762883e-06,
+      "loss": 0.1074,
+      "step": 5671
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.257324127635754e-06,
+      "loss": 0.1155,
+      "step": 5672
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.257190840508625e-06,
+      "loss": 0.1084,
+      "step": 5673
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.257057553381496e-06,
+      "loss": 0.1114,
+      "step": 5674
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.256924266254367e-06,
+      "loss": 0.095,
+      "step": 5675
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.256790979127238e-06,
+      "loss": 0.1588,
+      "step": 5676
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.256657692000107e-06,
+      "loss": 0.159,
+      "step": 5677
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.256524404872978e-06,
+      "loss": 0.1149,
+      "step": 5678
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.256391117745849e-06,
+      "loss": 0.1048,
+      "step": 5679
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.25625783061872e-06,
+      "loss": 0.0884,
+      "step": 5680
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.25612454349159e-06,
+      "loss": 0.103,
+      "step": 5681
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.255991256364462e-06,
+      "loss": 0.0963,
+      "step": 5682
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.255857969237331e-06,
+      "loss": 0.1499,
+      "step": 5683
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.255724682110202e-06,
+      "loss": 0.1192,
+      "step": 5684
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.255591394983073e-06,
+      "loss": 0.1033,
+      "step": 5685
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.255458107855944e-06,
+      "loss": 0.1057,
+      "step": 5686
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.255324820728815e-06,
+      "loss": 0.0909,
+      "step": 5687
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.255191533601684e-06,
+      "loss": 0.1088,
+      "step": 5688
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.255058246474557e-06,
+      "loss": 0.1183,
+      "step": 5689
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.254924959347426e-06,
+      "loss": 0.0714,
+      "step": 5690
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.254791672220297e-06,
+      "loss": 0.085,
+      "step": 5691
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.254658385093168e-06,
+      "loss": 0.1993,
+      "step": 5692
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.25452509796604e-06,
+      "loss": 0.1047,
+      "step": 5693
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.25439181083891e-06,
+      "loss": 0.1168,
+      "step": 5694
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.254258523711781e-06,
+      "loss": 0.1083,
+      "step": 5695
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.25412523658465e-06,
+      "loss": 0.1393,
+      "step": 5696
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.253991949457523e-06,
+      "loss": 0.088,
+      "step": 5697
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.253858662330393e-06,
+      "loss": 0.0812,
+      "step": 5698
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.253725375203264e-06,
+      "loss": 0.1367,
+      "step": 5699
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.253592088076135e-06,
+      "loss": 0.1472,
+      "step": 5700
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.253458800949006e-06,
+      "loss": 0.0793,
+      "step": 5701
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.253325513821877e-06,
+      "loss": 0.1355,
+      "step": 5702
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.253192226694746e-06,
+      "loss": 0.1168,
+      "step": 5703
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.253058939567617e-06,
+      "loss": 0.1276,
+      "step": 5704
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.252925652440488e-06,
+      "loss": 0.1291,
+      "step": 5705
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.252792365313359e-06,
+      "loss": 0.1134,
+      "step": 5706
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.25265907818623e-06,
+      "loss": 0.0916,
+      "step": 5707
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.252525791059101e-06,
+      "loss": 0.1141,
+      "step": 5708
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.25239250393197e-06,
+      "loss": 0.1023,
+      "step": 5709
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.252259216804841e-06,
+      "loss": 0.1018,
+      "step": 5710
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.252125929677712e-06,
+      "loss": 0.0665,
+      "step": 5711
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.251992642550583e-06,
+      "loss": 0.1097,
+      "step": 5712
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.251859355423454e-06,
+      "loss": 0.0977,
+      "step": 5713
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.251726068296325e-06,
+      "loss": 0.112,
+      "step": 5714
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.251592781169194e-06,
+      "loss": 0.1164,
+      "step": 5715
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.251459494042067e-06,
+      "loss": 0.0914,
+      "step": 5716
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.251326206914936e-06,
+      "loss": 0.1408,
+      "step": 5717
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.251192919787807e-06,
+      "loss": 0.1447,
+      "step": 5718
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.251059632660678e-06,
+      "loss": 0.1121,
+      "step": 5719
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.25092634553355e-06,
+      "loss": 0.1073,
+      "step": 5720
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.25079305840642e-06,
+      "loss": 0.0817,
+      "step": 5721
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.250659771279291e-06,
+      "loss": 0.1348,
+      "step": 5722
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.25052648415216e-06,
+      "loss": 0.0826,
+      "step": 5723
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.250393197025032e-06,
+      "loss": 0.102,
+      "step": 5724
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.250259909897903e-06,
+      "loss": 0.0833,
+      "step": 5725
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.250126622770774e-06,
+      "loss": 0.101,
+      "step": 5726
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.249993335643645e-06,
+      "loss": 0.114,
+      "step": 5727
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.249860048516514e-06,
+      "loss": 0.1033,
+      "step": 5728
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.249726761389387e-06,
+      "loss": 0.1706,
+      "step": 5729
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.249593474262256e-06,
+      "loss": 0.0911,
+      "step": 5730
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.249460187135127e-06,
+      "loss": 0.1416,
+      "step": 5731
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.249326900007998e-06,
+      "loss": 0.1064,
+      "step": 5732
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.249193612880869e-06,
+      "loss": 0.0853,
+      "step": 5733
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.24906032575374e-06,
+      "loss": 0.128,
+      "step": 5734
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.248927038626611e-06,
+      "loss": 0.1026,
+      "step": 5735
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.24879375149948e-06,
+      "loss": 0.1187,
+      "step": 5736
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.248660464372353e-06,
+      "loss": 0.0844,
+      "step": 5737
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.248527177245222e-06,
+      "loss": 0.1327,
+      "step": 5738
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.248393890118093e-06,
+      "loss": 0.135,
+      "step": 5739
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.248260602990964e-06,
+      "loss": 0.1279,
+      "step": 5740
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.248127315863835e-06,
+      "loss": 0.0814,
+      "step": 5741
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.247994028736706e-06,
+      "loss": 0.1477,
+      "step": 5742
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.247860741609576e-06,
+      "loss": 0.1051,
+      "step": 5743
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.247727454482447e-06,
+      "loss": 0.1128,
+      "step": 5744
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.247594167355318e-06,
+      "loss": 0.1086,
+      "step": 5745
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.247460880228189e-06,
+      "loss": 0.1221,
+      "step": 5746
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.24732759310106e-06,
+      "loss": 0.1013,
+      "step": 5747
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.24719430597393e-06,
+      "loss": 0.0906,
+      "step": 5748
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.2470610188468e-06,
+      "loss": 0.1179,
+      "step": 5749
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.24692773171967e-06,
+      "loss": 0.1195,
+      "step": 5750
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.246794444592542e-06,
+      "loss": 0.1085,
+      "step": 5751
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.246661157465413e-06,
+      "loss": 0.1452,
+      "step": 5752
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.246527870338284e-06,
+      "loss": 0.1091,
+      "step": 5753
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.246394583211155e-06,
+      "loss": 0.1059,
+      "step": 5754
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.246261296084024e-06,
+      "loss": 0.0782,
+      "step": 5755
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.246128008956897e-06,
+      "loss": 0.1201,
+      "step": 5756
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.245994721829766e-06,
+      "loss": 0.1093,
+      "step": 5757
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.245861434702637e-06,
+      "loss": 0.163,
+      "step": 5758
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.245728147575508e-06,
+      "loss": 0.1509,
+      "step": 5759
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.245594860448379e-06,
+      "loss": 0.1225,
+      "step": 5760
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.24546157332125e-06,
+      "loss": 0.1102,
+      "step": 5761
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.24532828619412e-06,
+      "loss": 0.1387,
+      "step": 5762
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.24519499906699e-06,
+      "loss": 0.0825,
+      "step": 5763
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.245061711939861e-06,
+      "loss": 0.1174,
+      "step": 5764
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.244928424812732e-06,
+      "loss": 0.1026,
+      "step": 5765
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.244795137685603e-06,
+      "loss": 0.0916,
+      "step": 5766
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.244661850558474e-06,
+      "loss": 0.096,
+      "step": 5767
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.244528563431344e-06,
+      "loss": 0.1044,
+      "step": 5768
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.244395276304216e-06,
+      "loss": 0.0865,
+      "step": 5769
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.244261989177086e-06,
+      "loss": 0.1177,
+      "step": 5770
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.244128702049957e-06,
+      "loss": 0.1571,
+      "step": 5771
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.243995414922828e-06,
+      "loss": 0.1286,
+      "step": 5772
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.243862127795699e-06,
+      "loss": 0.156,
+      "step": 5773
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.24372884066857e-06,
+      "loss": 0.0782,
+      "step": 5774
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.24359555354144e-06,
+      "loss": 0.0998,
+      "step": 5775
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.24346226641431e-06,
+      "loss": 0.1039,
+      "step": 5776
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.243328979287181e-06,
+      "loss": 0.1258,
+      "step": 5777
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.243195692160052e-06,
+      "loss": 0.0957,
+      "step": 5778
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.243062405032923e-06,
+      "loss": 0.1071,
+      "step": 5779
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.242929117905794e-06,
+      "loss": 0.1115,
+      "step": 5780
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.242795830778663e-06,
+      "loss": 0.1306,
+      "step": 5781
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.242662543651534e-06,
+      "loss": 0.1181,
+      "step": 5782
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.242529256524405e-06,
+      "loss": 0.1178,
+      "step": 5783
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.242395969397276e-06,
+      "loss": 0.1028,
+      "step": 5784
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.242262682270147e-06,
+      "loss": 0.1266,
+      "step": 5785
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.242129395143018e-06,
+      "loss": 0.1333,
+      "step": 5786
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.241996108015887e-06,
+      "loss": 0.1121,
+      "step": 5787
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.24186282088876e-06,
+      "loss": 0.1221,
+      "step": 5788
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.24172953376163e-06,
+      "loss": 0.1386,
+      "step": 5789
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.2415962466345e-06,
+      "loss": 0.1074,
+      "step": 5790
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.241462959507371e-06,
+      "loss": 0.149,
+      "step": 5791
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.241329672380242e-06,
+      "loss": 0.106,
+      "step": 5792
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.241196385253113e-06,
+      "loss": 0.1061,
+      "step": 5793
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.241063098125984e-06,
+      "loss": 0.0739,
+      "step": 5794
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.240929810998854e-06,
+      "loss": 0.1004,
+      "step": 5795
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.240796523871726e-06,
+      "loss": 0.1341,
+      "step": 5796
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.240663236744596e-06,
+      "loss": 0.0829,
+      "step": 5797
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.240529949617467e-06,
+      "loss": 0.121,
+      "step": 5798
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.240396662490338e-06,
+      "loss": 0.1081,
+      "step": 5799
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.240263375363209e-06,
+      "loss": 0.098,
+      "step": 5800
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.24013008823608e-06,
+      "loss": 0.0979,
+      "step": 5801
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.239996801108949e-06,
+      "loss": 0.1208,
+      "step": 5802
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.23986351398182e-06,
+      "loss": 0.0958,
+      "step": 5803
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.239730226854691e-06,
+      "loss": 0.0567,
+      "step": 5804
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.239596939727562e-06,
+      "loss": 0.1218,
+      "step": 5805
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.239463652600433e-06,
+      "loss": 0.0794,
+      "step": 5806
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.239330365473304e-06,
+      "loss": 0.1223,
+      "step": 5807
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.239197078346173e-06,
+      "loss": 0.1013,
+      "step": 5808
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.239063791219044e-06,
+      "loss": 0.1139,
+      "step": 5809
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.238930504091915e-06,
+      "loss": 0.1239,
+      "step": 5810
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.238797216964786e-06,
+      "loss": 0.0975,
+      "step": 5811
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.238663929837657e-06,
+      "loss": 0.1326,
+      "step": 5812
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.238530642710528e-06,
+      "loss": 0.1328,
+      "step": 5813
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.238397355583398e-06,
+      "loss": 0.129,
+      "step": 5814
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.23826406845627e-06,
+      "loss": 0.1882,
+      "step": 5815
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.23813078132914e-06,
+      "loss": 0.0948,
+      "step": 5816
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.23799749420201e-06,
+      "loss": 0.1056,
+      "step": 5817
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.237864207074881e-06,
+      "loss": 0.1181,
+      "step": 5818
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.237730919947752e-06,
+      "loss": 0.107,
+      "step": 5819
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.237597632820623e-06,
+      "loss": 0.0953,
+      "step": 5820
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.237464345693493e-06,
+      "loss": 0.0996,
+      "step": 5821
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.237331058566364e-06,
+      "loss": 0.0844,
+      "step": 5822
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.237197771439235e-06,
+      "loss": 0.0903,
+      "step": 5823
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.237064484312106e-06,
+      "loss": 0.0845,
+      "step": 5824
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.236931197184977e-06,
+      "loss": 0.1373,
+      "step": 5825
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.236797910057848e-06,
+      "loss": 0.1202,
+      "step": 5826
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.236664622930717e-06,
+      "loss": 0.1105,
+      "step": 5827
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.23653133580359e-06,
+      "loss": 0.137,
+      "step": 5828
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.236398048676459e-06,
+      "loss": 0.1133,
+      "step": 5829
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.23626476154933e-06,
+      "loss": 0.1054,
+      "step": 5830
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.236131474422201e-06,
+      "loss": 0.0894,
+      "step": 5831
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.235998187295072e-06,
+      "loss": 0.1108,
+      "step": 5832
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.235864900167943e-06,
+      "loss": 0.0982,
+      "step": 5833
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.235731613040814e-06,
+      "loss": 0.1046,
+      "step": 5834
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.235598325913683e-06,
+      "loss": 0.1178,
+      "step": 5835
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.235465038786556e-06,
+      "loss": 0.1485,
+      "step": 5836
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.235331751659425e-06,
+      "loss": 0.112,
+      "step": 5837
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.235198464532296e-06,
+      "loss": 0.175,
+      "step": 5838
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.235065177405167e-06,
+      "loss": 0.1131,
+      "step": 5839
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.234931890278037e-06,
+      "loss": 0.1055,
+      "step": 5840
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.23479860315091e-06,
+      "loss": 0.1345,
+      "step": 5841
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.234665316023779e-06,
+      "loss": 0.1044,
+      "step": 5842
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.23453202889665e-06,
+      "loss": 0.1344,
+      "step": 5843
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.23439874176952e-06,
+      "loss": 0.1449,
+      "step": 5844
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.234265454642392e-06,
+      "loss": 0.089,
+      "step": 5845
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.234132167515263e-06,
+      "loss": 0.131,
+      "step": 5846
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.233998880388134e-06,
+      "loss": 0.102,
+      "step": 5847
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.233865593261003e-06,
+      "loss": 0.1284,
+      "step": 5848
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.233732306133874e-06,
+      "loss": 0.1078,
+      "step": 5849
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.233599019006745e-06,
+      "loss": 0.1216,
+      "step": 5850
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.233465731879616e-06,
+      "loss": 0.1359,
+      "step": 5851
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.233332444752487e-06,
+      "loss": 0.152,
+      "step": 5852
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.233199157625358e-06,
+      "loss": 0.1295,
+      "step": 5853
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.233065870498227e-06,
+      "loss": 0.0873,
+      "step": 5854
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.2329325833711e-06,
+      "loss": 0.1088,
+      "step": 5855
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.232799296243969e-06,
+      "loss": 0.0954,
+      "step": 5856
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.23266600911684e-06,
+      "loss": 0.0877,
+      "step": 5857
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.232532721989711e-06,
+      "loss": 0.1331,
+      "step": 5858
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.232399434862582e-06,
+      "loss": 0.1209,
+      "step": 5859
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.232266147735453e-06,
+      "loss": 0.1145,
+      "step": 5860
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.232132860608322e-06,
+      "loss": 0.074,
+      "step": 5861
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.231999573481193e-06,
+      "loss": 0.0996,
+      "step": 5862
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.231866286354064e-06,
+      "loss": 0.1011,
+      "step": 5863
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.231732999226935e-06,
+      "loss": 0.0849,
+      "step": 5864
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.231599712099806e-06,
+      "loss": 0.1772,
+      "step": 5865
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.231466424972677e-06,
+      "loss": 0.1334,
+      "step": 5866
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.231333137845547e-06,
+      "loss": 0.0925,
+      "step": 5867
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.23119985071842e-06,
+      "loss": 0.1622,
+      "step": 5868
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.231066563591289e-06,
+      "loss": 0.1166,
+      "step": 5869
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.23093327646416e-06,
+      "loss": 0.164,
+      "step": 5870
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.23079998933703e-06,
+      "loss": 0.1198,
+      "step": 5871
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.230666702209902e-06,
+      "loss": 0.128,
+      "step": 5872
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.230533415082773e-06,
+      "loss": 0.1035,
+      "step": 5873
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.230400127955644e-06,
+      "loss": 0.1177,
+      "step": 5874
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.230266840828513e-06,
+      "loss": 0.1299,
+      "step": 5875
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.230133553701384e-06,
+      "loss": 0.1269,
+      "step": 5876
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.230000266574255e-06,
+      "loss": 0.1297,
+      "step": 5877
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.229866979447126e-06,
+      "loss": 0.1341,
+      "step": 5878
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.229733692319997e-06,
+      "loss": 0.1644,
+      "step": 5879
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.229600405192866e-06,
+      "loss": 0.0999,
+      "step": 5880
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.229467118065737e-06,
+      "loss": 0.0735,
+      "step": 5881
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.229333830938608e-06,
+      "loss": 0.0837,
+      "step": 5882
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.22920054381148e-06,
+      "loss": 0.1279,
+      "step": 5883
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.22906725668435e-06,
+      "loss": 0.1182,
+      "step": 5884
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 9.228933969557221e-06,
+      "loss": 0.0967,
+      "step": 5885
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 9.22880068243009e-06,
+      "loss": 0.1075,
+      "step": 5886
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 9.228667395302963e-06,
+      "loss": 0.1334,
+      "step": 5887
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 9.228534108175832e-06,
+      "loss": 0.1123,
+      "step": 5888
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 9.228400821048703e-06,
+      "loss": 0.0722,
+      "step": 5889
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 9.228267533921574e-06,
+      "loss": 0.1126,
+      "step": 5890
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 9.228134246794445e-06,
+      "loss": 0.1298,
+      "step": 5891
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 9.228000959667316e-06,
+      "loss": 0.1285,
+      "step": 5892
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 9.227867672540187e-06,
+      "loss": 0.1268,
+      "step": 5893
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 9.227734385413057e-06,
+      "loss": 0.1167,
+      "step": 5894
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 9.22760109828593e-06,
+      "loss": 0.0772,
+      "step": 5895
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 9.227467811158799e-06,
+      "loss": 0.107,
+      "step": 5896
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 9.22733452403167e-06,
+      "loss": 0.0872,
+      "step": 5897
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 9.22720123690454e-06,
+      "loss": 0.1646,
+      "step": 5898
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 9.22706794977741e-06,
+      "loss": 0.1247,
+      "step": 5899
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 9.226934662650283e-06,
+      "loss": 0.1085,
+      "step": 5900
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 9.226801375523152e-06,
+      "loss": 0.1259,
+      "step": 5901
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 9.226668088396023e-06,
+      "loss": 0.1118,
+      "step": 5902
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 9.226534801268894e-06,
+      "loss": 0.1163,
+      "step": 5903
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 9.226401514141765e-06,
+      "loss": 0.0968,
+      "step": 5904
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 9.226268227014636e-06,
+      "loss": 0.0751,
+      "step": 5905
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 9.226134939887507e-06,
+      "loss": 0.1232,
+      "step": 5906
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 9.226001652760376e-06,
+      "loss": 0.0955,
+      "step": 5907
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 9.225868365633249e-06,
+      "loss": 0.1597,
+      "step": 5908
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 9.225735078506118e-06,
+      "loss": 0.1154,
+      "step": 5909
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 9.22560179137899e-06,
+      "loss": 0.135,
+      "step": 5910
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 9.22546850425186e-06,
+      "loss": 0.1026,
+      "step": 5911
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 9.225335217124731e-06,
+      "loss": 0.1281,
+      "step": 5912
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 9.225201929997602e-06,
+      "loss": 0.1961,
+      "step": 5913
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 9.225068642870473e-06,
+      "loss": 0.1361,
+      "step": 5914
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 9.224935355743343e-06,
+      "loss": 0.0807,
+      "step": 5915
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 9.224802068616214e-06,
+      "loss": 0.1486,
+      "step": 5916
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 9.224668781489085e-06,
+      "loss": 0.1898,
+      "step": 5917
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 9.224535494361956e-06,
+      "loss": 0.1178,
+      "step": 5918
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 9.224402207234827e-06,
+      "loss": 0.1359,
+      "step": 5919
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 9.224268920107696e-06,
+      "loss": 0.1055,
+      "step": 5920
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 9.224135632980567e-06,
+      "loss": 0.0882,
+      "step": 5921
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 9.224002345853438e-06,
+      "loss": 0.1195,
+      "step": 5922
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 9.223869058726309e-06,
+      "loss": 0.0848,
+      "step": 5923
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 9.22373577159918e-06,
+      "loss": 0.1326,
+      "step": 5924
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 9.22360248447205e-06,
+      "loss": 0.1306,
+      "step": 5925
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 9.22346919734492e-06,
+      "loss": 0.1427,
+      "step": 5926
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 9.223335910217793e-06,
+      "loss": 0.0888,
+      "step": 5927
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 9.223202623090662e-06,
+      "loss": 0.1055,
+      "step": 5928
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 9.223069335963533e-06,
+      "loss": 0.091,
+      "step": 5929
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 9.222936048836404e-06,
+      "loss": 0.0867,
+      "step": 5930
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 9.222802761709275e-06,
+      "loss": 0.0833,
+      "step": 5931
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 9.222669474582146e-06,
+      "loss": 0.0864,
+      "step": 5932
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 9.222536187455017e-06,
+      "loss": 0.1077,
+      "step": 5933
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 9.222402900327886e-06,
+      "loss": 0.098,
+      "step": 5934
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 9.222269613200759e-06,
+      "loss": 0.0853,
+      "step": 5935
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 9.222136326073628e-06,
+      "loss": 0.1151,
+      "step": 5936
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 9.2220030389465e-06,
+      "loss": 0.0961,
+      "step": 5937
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 9.22186975181937e-06,
+      "loss": 0.1215,
+      "step": 5938
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 9.22173646469224e-06,
+      "loss": 0.1235,
+      "step": 5939
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 9.221603177565112e-06,
+      "loss": 0.0859,
+      "step": 5940
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 9.221469890437982e-06,
+      "loss": 0.1308,
+      "step": 5941
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 9.221336603310853e-06,
+      "loss": 0.1233,
+      "step": 5942
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 9.221203316183724e-06,
+      "loss": 0.122,
+      "step": 5943
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 9.221070029056595e-06,
+      "loss": 0.0945,
+      "step": 5944
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 9.220936741929466e-06,
+      "loss": 0.1204,
+      "step": 5945
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 9.220803454802337e-06,
+      "loss": 0.1177,
+      "step": 5946
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 9.220670167675206e-06,
+      "loss": 0.1381,
+      "step": 5947
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 9.220536880548077e-06,
+      "loss": 0.0811,
+      "step": 5948
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 9.220403593420948e-06,
+      "loss": 0.0873,
+      "step": 5949
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 9.220270306293819e-06,
+      "loss": 0.1142,
+      "step": 5950
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 9.22013701916669e-06,
+      "loss": 0.1366,
+      "step": 5951
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 9.220003732039561e-06,
+      "loss": 0.1171,
+      "step": 5952
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 9.21987044491243e-06,
+      "loss": 0.1142,
+      "step": 5953
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 9.219737157785303e-06,
+      "loss": 0.1521,
+      "step": 5954
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 9.219603870658172e-06,
+      "loss": 0.119,
+      "step": 5955
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 9.219470583531043e-06,
+      "loss": 0.132,
+      "step": 5956
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 9.219337296403914e-06,
+      "loss": 0.0982,
+      "step": 5957
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 9.219204009276783e-06,
+      "loss": 0.1425,
+      "step": 5958
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 9.219070722149656e-06,
+      "loss": 0.111,
+      "step": 5959
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 9.218937435022525e-06,
+      "loss": 0.1103,
+      "step": 5960
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 9.218804147895396e-06,
+      "loss": 0.0783,
+      "step": 5961
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 9.218670860768267e-06,
+      "loss": 0.1076,
+      "step": 5962
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 9.218537573641138e-06,
+      "loss": 0.131,
+      "step": 5963
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 9.21840428651401e-06,
+      "loss": 0.1177,
+      "step": 5964
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 9.21827099938688e-06,
+      "loss": 0.0917,
+      "step": 5965
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 9.21813771225975e-06,
+      "loss": 0.0983,
+      "step": 5966
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 9.218004425132622e-06,
+      "loss": 0.1191,
+      "step": 5967
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 9.217871138005492e-06,
+      "loss": 0.105,
+      "step": 5968
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 9.217737850878363e-06,
+      "loss": 0.1467,
+      "step": 5969
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 9.217604563751234e-06,
+      "loss": 0.095,
+      "step": 5970
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 9.217471276624105e-06,
+      "loss": 0.0904,
+      "step": 5971
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 9.217337989496976e-06,
+      "loss": 0.0978,
+      "step": 5972
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 9.217204702369847e-06,
+      "loss": 0.151,
+      "step": 5973
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 9.217071415242716e-06,
+      "loss": 0.1406,
+      "step": 5974
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 9.216938128115589e-06,
+      "loss": 0.0842,
+      "step": 5975
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 9.216804840988458e-06,
+      "loss": 0.0954,
+      "step": 5976
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 9.216671553861329e-06,
+      "loss": 0.1453,
+      "step": 5977
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 9.2165382667342e-06,
+      "loss": 0.1462,
+      "step": 5978
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 9.21640497960707e-06,
+      "loss": 0.1297,
+      "step": 5979
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 9.216271692479942e-06,
+      "loss": 0.1359,
+      "step": 5980
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 9.216138405352811e-06,
+      "loss": 0.1407,
+      "step": 5981
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 9.216005118225682e-06,
+      "loss": 0.1006,
+      "step": 5982
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 9.215871831098553e-06,
+      "loss": 0.1347,
+      "step": 5983
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 9.215738543971424e-06,
+      "loss": 0.1264,
+      "step": 5984
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 9.215605256844295e-06,
+      "loss": 0.1096,
+      "step": 5985
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 9.215471969717166e-06,
+      "loss": 0.1352,
+      "step": 5986
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 9.215338682590036e-06,
+      "loss": 0.0858,
+      "step": 5987
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 9.215205395462907e-06,
+      "loss": 0.1241,
+      "step": 5988
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 9.215072108335778e-06,
+      "loss": 0.1072,
+      "step": 5989
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 9.214938821208649e-06,
+      "loss": 0.1077,
+      "step": 5990
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 9.21480553408152e-06,
+      "loss": 0.0853,
+      "step": 5991
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 9.21467224695439e-06,
+      "loss": 0.1042,
+      "step": 5992
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 9.21453895982726e-06,
+      "loss": 0.1278,
+      "step": 5993
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 9.214405672700133e-06,
+      "loss": 0.0986,
+      "step": 5994
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 9.214272385573002e-06,
+      "loss": 0.1163,
+      "step": 5995
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 9.214139098445873e-06,
+      "loss": 0.1462,
+      "step": 5996
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 9.214005811318744e-06,
+      "loss": 0.1131,
+      "step": 5997
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 9.213872524191613e-06,
+      "loss": 0.0925,
+      "step": 5998
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 9.213739237064486e-06,
+      "loss": 0.1135,
+      "step": 5999
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 9.213605949937355e-06,
+      "loss": 0.1286,
+      "step": 6000
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 9.213472662810226e-06,
+      "loss": 0.1237,
+      "step": 6001
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 9.213339375683097e-06,
+      "loss": 0.0933,
+      "step": 6002
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 9.213206088555968e-06,
+      "loss": 0.0834,
+      "step": 6003
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 9.213072801428839e-06,
+      "loss": 0.1801,
+      "step": 6004
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 9.21293951430171e-06,
+      "loss": 0.1416,
+      "step": 6005
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 9.21280622717458e-06,
+      "loss": 0.1543,
+      "step": 6006
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 9.212672940047452e-06,
+      "loss": 0.089,
+      "step": 6007
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 9.212539652920321e-06,
+      "loss": 0.1054,
+      "step": 6008
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 9.212406365793192e-06,
+      "loss": 0.1426,
+      "step": 6009
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 9.212273078666063e-06,
+      "loss": 0.0736,
+      "step": 6010
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 9.212139791538934e-06,
+      "loss": 0.1148,
+      "step": 6011
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 9.212006504411805e-06,
+      "loss": 0.1249,
+      "step": 6012
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 9.211873217284676e-06,
+      "loss": 0.1074,
+      "step": 6013
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 9.211739930157546e-06,
+      "loss": 0.0863,
+      "step": 6014
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 9.211606643030417e-06,
+      "loss": 0.1436,
+      "step": 6015
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 9.211473355903288e-06,
+      "loss": 0.1093,
+      "step": 6016
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 9.211340068776159e-06,
+      "loss": 0.1164,
+      "step": 6017
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 9.21120678164903e-06,
+      "loss": 0.0891,
+      "step": 6018
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 9.211073494521899e-06,
+      "loss": 0.1006,
+      "step": 6019
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 9.21094020739477e-06,
+      "loss": 0.0974,
+      "step": 6020
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 9.210806920267641e-06,
+      "loss": 0.105,
+      "step": 6021
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 9.210673633140512e-06,
+      "loss": 0.1242,
+      "step": 6022
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 9.210540346013383e-06,
+      "loss": 0.109,
+      "step": 6023
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 9.210407058886254e-06,
+      "loss": 0.1593,
+      "step": 6024
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 9.210273771759123e-06,
+      "loss": 0.1154,
+      "step": 6025
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 9.210140484631996e-06,
+      "loss": 0.1324,
+      "step": 6026
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 9.210007197504865e-06,
+      "loss": 0.1063,
+      "step": 6027
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 9.209873910377736e-06,
+      "loss": 0.1513,
+      "step": 6028
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 9.209740623250607e-06,
+      "loss": 0.0638,
+      "step": 6029
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 9.209607336123478e-06,
+      "loss": 0.1154,
+      "step": 6030
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 9.209474048996349e-06,
+      "loss": 0.1042,
+      "step": 6031
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 9.20934076186922e-06,
+      "loss": 0.1179,
+      "step": 6032
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 9.20920747474209e-06,
+      "loss": 0.1123,
+      "step": 6033
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 9.209074187614962e-06,
+      "loss": 0.1024,
+      "step": 6034
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 9.208940900487831e-06,
+      "loss": 0.0868,
+      "step": 6035
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 9.208807613360702e-06,
+      "loss": 0.1087,
+      "step": 6036
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 9.208674326233573e-06,
+      "loss": 0.1279,
+      "step": 6037
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 9.208541039106443e-06,
+      "loss": 0.1092,
+      "step": 6038
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 9.208407751979315e-06,
+      "loss": 0.1046,
+      "step": 6039
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 9.208274464852185e-06,
+      "loss": 0.0943,
+      "step": 6040
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 9.208141177725056e-06,
+      "loss": 0.0911,
+      "step": 6041
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 9.208007890597927e-06,
+      "loss": 0.1041,
+      "step": 6042
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 9.207874603470798e-06,
+      "loss": 0.1932,
+      "step": 6043
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 9.207741316343669e-06,
+      "loss": 0.1169,
+      "step": 6044
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 9.20760802921654e-06,
+      "loss": 0.1233,
+      "step": 6045
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 9.207474742089409e-06,
+      "loss": 0.1253,
+      "step": 6046
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 9.20734145496228e-06,
+      "loss": 0.111,
+      "step": 6047
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 9.207208167835151e-06,
+      "loss": 0.0982,
+      "step": 6048
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 9.207074880708022e-06,
+      "loss": 0.1474,
+      "step": 6049
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 9.206941593580893e-06,
+      "loss": 0.086,
+      "step": 6050
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 9.206808306453764e-06,
+      "loss": 0.0684,
+      "step": 6051
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 9.206675019326633e-06,
+      "loss": 0.0836,
+      "step": 6052
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 9.206541732199506e-06,
+      "loss": 0.1058,
+      "step": 6053
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 9.206408445072375e-06,
+      "loss": 0.1274,
+      "step": 6054
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 9.206275157945246e-06,
+      "loss": 0.1306,
+      "step": 6055
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 9.206141870818117e-06,
+      "loss": 0.1154,
+      "step": 6056
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 9.206008583690987e-06,
+      "loss": 0.1023,
+      "step": 6057
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 9.20587529656386e-06,
+      "loss": 0.1161,
+      "step": 6058
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 9.205742009436729e-06,
+      "loss": 0.1317,
+      "step": 6059
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 9.2056087223096e-06,
+      "loss": 0.1042,
+      "step": 6060
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 9.20547543518247e-06,
+      "loss": 0.103,
+      "step": 6061
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 9.205342148055342e-06,
+      "loss": 0.127,
+      "step": 6062
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 9.205208860928213e-06,
+      "loss": 0.0914,
+      "step": 6063
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 9.205075573801084e-06,
+      "loss": 0.1036,
+      "step": 6064
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 9.204942286673953e-06,
+      "loss": 0.1444,
+      "step": 6065
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 9.204808999546825e-06,
+      "loss": 0.1128,
+      "step": 6066
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 9.204675712419695e-06,
+      "loss": 0.125,
+      "step": 6067
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 9.204542425292566e-06,
+      "loss": 0.1004,
+      "step": 6068
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 9.204409138165437e-06,
+      "loss": 0.1259,
+      "step": 6069
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 9.204275851038308e-06,
+      "loss": 0.1082,
+      "step": 6070
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 9.204142563911179e-06,
+      "loss": 0.1407,
+      "step": 6071
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 9.20400927678405e-06,
+      "loss": 0.0984,
+      "step": 6072
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 9.203875989656919e-06,
+      "loss": 0.1281,
+      "step": 6073
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 9.203742702529792e-06,
+      "loss": 0.0996,
+      "step": 6074
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 9.203609415402661e-06,
+      "loss": 0.0953,
+      "step": 6075
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 9.203476128275532e-06,
+      "loss": 0.1416,
+      "step": 6076
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 9.203342841148403e-06,
+      "loss": 0.1047,
+      "step": 6077
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 9.203209554021272e-06,
+      "loss": 0.1663,
+      "step": 6078
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 9.203076266894145e-06,
+      "loss": 0.0867,
+      "step": 6079
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 9.202942979767014e-06,
+      "loss": 0.1063,
+      "step": 6080
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 9.202809692639885e-06,
+      "loss": 0.1446,
+      "step": 6081
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 9.202676405512756e-06,
+      "loss": 0.1002,
+      "step": 6082
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 9.202543118385627e-06,
+      "loss": 0.1041,
+      "step": 6083
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 9.202409831258498e-06,
+      "loss": 0.1058,
+      "step": 6084
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 9.20227654413137e-06,
+      "loss": 0.1225,
+      "step": 6085
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 9.202143257004239e-06,
+      "loss": 0.1091,
+      "step": 6086
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 9.20200996987711e-06,
+      "loss": 0.1069,
+      "step": 6087
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 9.20187668274998e-06,
+      "loss": 0.0869,
+      "step": 6088
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 9.201743395622852e-06,
+      "loss": 0.0809,
+      "step": 6089
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 9.201610108495723e-06,
+      "loss": 0.1145,
+      "step": 6090
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 9.201476821368594e-06,
+      "loss": 0.1475,
+      "step": 6091
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 9.201343534241463e-06,
+      "loss": 0.1165,
+      "step": 6092
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 9.201210247114336e-06,
+      "loss": 0.1307,
+      "step": 6093
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 9.201076959987205e-06,
+      "loss": 0.1533,
+      "step": 6094
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 9.200943672860076e-06,
+      "loss": 0.1001,
+      "step": 6095
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 9.200810385732947e-06,
+      "loss": 0.0881,
+      "step": 6096
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 9.200677098605816e-06,
+      "loss": 0.1623,
+      "step": 6097
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 9.200543811478689e-06,
+      "loss": 0.1167,
+      "step": 6098
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 9.200410524351558e-06,
+      "loss": 0.1438,
+      "step": 6099
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 9.200277237224429e-06,
+      "loss": 0.0793,
+      "step": 6100
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 9.2001439500973e-06,
+      "loss": 0.1286,
+      "step": 6101
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 9.200010662970171e-06,
+      "loss": 0.1199,
+      "step": 6102
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 9.199877375843042e-06,
+      "loss": 0.0759,
+      "step": 6103
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 9.199744088715913e-06,
+      "loss": 0.1325,
+      "step": 6104
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 9.199610801588782e-06,
+      "loss": 0.1402,
+      "step": 6105
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 9.199477514461655e-06,
+      "loss": 0.1131,
+      "step": 6106
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 9.199344227334524e-06,
+      "loss": 0.1002,
+      "step": 6107
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 9.199210940207395e-06,
+      "loss": 0.0854,
+      "step": 6108
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 9.199077653080266e-06,
+      "loss": 0.0828,
+      "step": 6109
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 9.198944365953137e-06,
+      "loss": 0.0999,
+      "step": 6110
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 9.198811078826008e-06,
+      "loss": 0.0937,
+      "step": 6111
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 9.19867779169888e-06,
+      "loss": 0.1157,
+      "step": 6112
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 9.198544504571749e-06,
+      "loss": 0.0794,
+      "step": 6113
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 9.19841121744462e-06,
+      "loss": 0.132,
+      "step": 6114
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 9.19827793031749e-06,
+      "loss": 0.1178,
+      "step": 6115
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 9.198144643190362e-06,
+      "loss": 0.0971,
+      "step": 6116
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 9.198011356063233e-06,
+      "loss": 0.1081,
+      "step": 6117
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 9.197878068936102e-06,
+      "loss": 0.1081,
+      "step": 6118
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 9.197744781808973e-06,
+      "loss": 0.0929,
+      "step": 6119
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 9.197611494681844e-06,
+      "loss": 0.1424,
+      "step": 6120
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 9.197478207554715e-06,
+      "loss": 0.1249,
+      "step": 6121
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 9.197344920427586e-06,
+      "loss": 0.1298,
+      "step": 6122
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 9.197211633300457e-06,
+      "loss": 0.1416,
+      "step": 6123
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 9.197078346173326e-06,
+      "loss": 0.0946,
+      "step": 6124
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 9.196945059046199e-06,
+      "loss": 0.0954,
+      "step": 6125
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 9.196811771919068e-06,
+      "loss": 0.0994,
+      "step": 6126
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 9.19667848479194e-06,
+      "loss": 0.1047,
+      "step": 6127
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 9.19654519766481e-06,
+      "loss": 0.1432,
+      "step": 6128
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 9.196411910537681e-06,
+      "loss": 0.1149,
+      "step": 6129
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 9.196278623410552e-06,
+      "loss": 0.0588,
+      "step": 6130
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 9.196145336283423e-06,
+      "loss": 0.1201,
+      "step": 6131
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 9.196012049156293e-06,
+      "loss": 0.1756,
+      "step": 6132
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 9.195878762029165e-06,
+      "loss": 0.1432,
+      "step": 6133
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 9.195745474902035e-06,
+      "loss": 0.0912,
+      "step": 6134
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 9.195612187774905e-06,
+      "loss": 0.0958,
+      "step": 6135
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 9.195478900647776e-06,
+      "loss": 0.1018,
+      "step": 6136
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 9.195345613520646e-06,
+      "loss": 0.1401,
+      "step": 6137
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 9.195212326393518e-06,
+      "loss": 0.1336,
+      "step": 6138
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 9.195079039266388e-06,
+      "loss": 0.0761,
+      "step": 6139
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 9.194945752139259e-06,
+      "loss": 0.116,
+      "step": 6140
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 9.19481246501213e-06,
+      "loss": 0.113,
+      "step": 6141
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 9.194679177885e-06,
+      "loss": 0.1198,
+      "step": 6142
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 9.194545890757872e-06,
+      "loss": 0.0801,
+      "step": 6143
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 9.194412603630743e-06,
+      "loss": 0.1173,
+      "step": 6144
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 9.194279316503612e-06,
+      "loss": 0.082,
+      "step": 6145
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 9.194146029376485e-06,
+      "loss": 0.0831,
+      "step": 6146
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 9.194012742249354e-06,
+      "loss": 0.1247,
+      "step": 6147
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 9.193879455122225e-06,
+      "loss": 0.1243,
+      "step": 6148
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 9.193746167995096e-06,
+      "loss": 0.0915,
+      "step": 6149
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 9.193612880867967e-06,
+      "loss": 0.107,
+      "step": 6150
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 9.193479593740838e-06,
+      "loss": 0.0963,
+      "step": 6151
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 9.193346306613709e-06,
+      "loss": 0.0868,
+      "step": 6152
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 9.193213019486578e-06,
+      "loss": 0.1457,
+      "step": 6153
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 9.19307973235945e-06,
+      "loss": 0.0963,
+      "step": 6154
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 9.19294644523232e-06,
+      "loss": 0.1154,
+      "step": 6155
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 9.192813158105191e-06,
+      "loss": 0.0948,
+      "step": 6156
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 9.192679870978062e-06,
+      "loss": 0.0994,
+      "step": 6157
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 9.192546583850932e-06,
+      "loss": 0.0763,
+      "step": 6158
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 9.192413296723803e-06,
+      "loss": 0.1014,
+      "step": 6159
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 9.192280009596674e-06,
+      "loss": 0.0945,
+      "step": 6160
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 9.192146722469545e-06,
+      "loss": 0.0985,
+      "step": 6161
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 9.192013435342416e-06,
+      "loss": 0.1085,
+      "step": 6162
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 9.191880148215287e-06,
+      "loss": 0.1102,
+      "step": 6163
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 9.191746861088156e-06,
+      "loss": 0.091,
+      "step": 6164
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 9.191613573961029e-06,
+      "loss": 0.068,
+      "step": 6165
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 9.191480286833898e-06,
+      "loss": 0.1116,
+      "step": 6166
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 9.191346999706769e-06,
+      "loss": 0.0782,
+      "step": 6167
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 9.19121371257964e-06,
+      "loss": 0.1199,
+      "step": 6168
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 9.19108042545251e-06,
+      "loss": 0.0884,
+      "step": 6169
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 9.190947138325382e-06,
+      "loss": 0.085,
+      "step": 6170
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 9.190813851198253e-06,
+      "loss": 0.1085,
+      "step": 6171
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 9.190680564071122e-06,
+      "loss": 0.1206,
+      "step": 6172
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 9.190547276943993e-06,
+      "loss": 0.1064,
+      "step": 6173
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 9.190413989816864e-06,
+      "loss": 0.1195,
+      "step": 6174
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 9.190280702689735e-06,
+      "loss": 0.1068,
+      "step": 6175
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 9.190147415562606e-06,
+      "loss": 0.1129,
+      "step": 6176
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 9.190014128435475e-06,
+      "loss": 0.1168,
+      "step": 6177
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 9.189880841308348e-06,
+      "loss": 0.089,
+      "step": 6178
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 9.189747554181217e-06,
+      "loss": 0.0996,
+      "step": 6179
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 9.189614267054088e-06,
+      "loss": 0.1063,
+      "step": 6180
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 9.18948097992696e-06,
+      "loss": 0.1054,
+      "step": 6181
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 9.18934769279983e-06,
+      "loss": 0.0963,
+      "step": 6182
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 9.189214405672701e-06,
+      "loss": 0.0784,
+      "step": 6183
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 9.189081118545572e-06,
+      "loss": 0.1292,
+      "step": 6184
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 9.188947831418442e-06,
+      "loss": 0.1505,
+      "step": 6185
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 9.188814544291313e-06,
+      "loss": 0.1151,
+      "step": 6186
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 9.188681257164184e-06,
+      "loss": 0.0969,
+      "step": 6187
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 9.188547970037055e-06,
+      "loss": 0.0769,
+      "step": 6188
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 9.188414682909926e-06,
+      "loss": 0.1013,
+      "step": 6189
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 9.188281395782797e-06,
+      "loss": 0.1028,
+      "step": 6190
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 9.188148108655666e-06,
+      "loss": 0.0863,
+      "step": 6191
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 9.188014821528539e-06,
+      "loss": 0.1369,
+      "step": 6192
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 9.187881534401408e-06,
+      "loss": 0.1363,
+      "step": 6193
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 9.187748247274279e-06,
+      "loss": 0.0923,
+      "step": 6194
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 9.18761496014715e-06,
+      "loss": 0.0878,
+      "step": 6195
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 9.18748167302002e-06,
+      "loss": 0.1294,
+      "step": 6196
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 9.187348385892892e-06,
+      "loss": 0.1516,
+      "step": 6197
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 9.187215098765761e-06,
+      "loss": 0.1409,
+      "step": 6198
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 9.187081811638632e-06,
+      "loss": 0.0873,
+      "step": 6199
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 9.186948524511503e-06,
+      "loss": 0.0876,
+      "step": 6200
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 9.186815237384374e-06,
+      "loss": 0.0837,
+      "step": 6201
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 9.186681950257245e-06,
+      "loss": 0.1033,
+      "step": 6202
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 9.186548663130116e-06,
+      "loss": 0.0999,
+      "step": 6203
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 9.186415376002985e-06,
+      "loss": 0.1276,
+      "step": 6204
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 9.186282088875858e-06,
+      "loss": 0.1093,
+      "step": 6205
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 9.186148801748727e-06,
+      "loss": 0.0887,
+      "step": 6206
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 9.186015514621598e-06,
+      "loss": 0.0924,
+      "step": 6207
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 9.18588222749447e-06,
+      "loss": 0.096,
+      "step": 6208
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 9.18574894036734e-06,
+      "loss": 0.0855,
+      "step": 6209
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 9.185615653240211e-06,
+      "loss": 0.0983,
+      "step": 6210
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 9.185482366113082e-06,
+      "loss": 0.1054,
+      "step": 6211
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 9.185349078985952e-06,
+      "loss": 0.0833,
+      "step": 6212
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 9.185215791858823e-06,
+      "loss": 0.0816,
+      "step": 6213
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 9.185082504731694e-06,
+      "loss": 0.1192,
+      "step": 6214
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 9.184949217604565e-06,
+      "loss": 0.1309,
+      "step": 6215
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 9.184815930477436e-06,
+      "loss": 0.1323,
+      "step": 6216
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 9.184682643350305e-06,
+      "loss": 0.1039,
+      "step": 6217
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 9.184549356223178e-06,
+      "loss": 0.145,
+      "step": 6218
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 9.184416069096047e-06,
+      "loss": 0.0717,
+      "step": 6219
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 9.184282781968918e-06,
+      "loss": 0.1443,
+      "step": 6220
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 9.184149494841789e-06,
+      "loss": 0.1275,
+      "step": 6221
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 9.18401620771466e-06,
+      "loss": 0.1716,
+      "step": 6222
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 9.183882920587531e-06,
+      "loss": 0.1095,
+      "step": 6223
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 9.183749633460402e-06,
+      "loss": 0.1281,
+      "step": 6224
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 9.183616346333271e-06,
+      "loss": 0.1007,
+      "step": 6225
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 9.183483059206142e-06,
+      "loss": 0.0887,
+      "step": 6226
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 9.183349772079013e-06,
+      "loss": 0.1105,
+      "step": 6227
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 9.183216484951884e-06,
+      "loss": 0.0922,
+      "step": 6228
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 9.183083197824755e-06,
+      "loss": 0.1469,
+      "step": 6229
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 9.182949910697626e-06,
+      "loss": 0.1556,
+      "step": 6230
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 9.182816623570496e-06,
+      "loss": 0.1088,
+      "step": 6231
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 9.182683336443367e-06,
+      "loss": 0.0786,
+      "step": 6232
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 9.182550049316238e-06,
+      "loss": 0.0863,
+      "step": 6233
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 9.182416762189109e-06,
+      "loss": 0.1067,
+      "step": 6234
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 9.18228347506198e-06,
+      "loss": 0.1226,
+      "step": 6235
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 9.182150187934849e-06,
+      "loss": 0.1258,
+      "step": 6236
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 9.182016900807722e-06,
+      "loss": 0.0934,
+      "step": 6237
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 9.18188361368059e-06,
+      "loss": 0.1268,
+      "step": 6238
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 9.181750326553462e-06,
+      "loss": 0.1146,
+      "step": 6239
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 9.181617039426333e-06,
+      "loss": 0.1003,
+      "step": 6240
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 9.181483752299204e-06,
+      "loss": 0.1161,
+      "step": 6241
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 9.181350465172075e-06,
+      "loss": 0.0823,
+      "step": 6242
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 9.181217178044946e-06,
+      "loss": 0.1016,
+      "step": 6243
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 9.181083890917815e-06,
+      "loss": 0.0682,
+      "step": 6244
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 9.180950603790688e-06,
+      "loss": 0.1134,
+      "step": 6245
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 9.180817316663557e-06,
+      "loss": 0.1113,
+      "step": 6246
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 9.180684029536428e-06,
+      "loss": 0.1166,
+      "step": 6247
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 9.180550742409299e-06,
+      "loss": 0.1459,
+      "step": 6248
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 9.18041745528217e-06,
+      "loss": 0.1347,
+      "step": 6249
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 9.180284168155041e-06,
+      "loss": 0.1019,
+      "step": 6250
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 9.18015088102791e-06,
+      "loss": 0.1232,
+      "step": 6251
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 9.180017593900781e-06,
+      "loss": 0.0876,
+      "step": 6252
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 9.179884306773652e-06,
+      "loss": 0.1025,
+      "step": 6253
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 9.179751019646523e-06,
+      "loss": 0.0902,
+      "step": 6254
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 9.179617732519394e-06,
+      "loss": 0.1157,
+      "step": 6255
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 9.179484445392265e-06,
+      "loss": 0.1202,
+      "step": 6256
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 9.179351158265135e-06,
+      "loss": 0.079,
+      "step": 6257
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 9.179217871138006e-06,
+      "loss": 0.1292,
+      "step": 6258
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 9.179084584010877e-06,
+      "loss": 0.1186,
+      "step": 6259
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 9.178951296883748e-06,
+      "loss": 0.1612,
+      "step": 6260
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 9.178818009756619e-06,
+      "loss": 0.0732,
+      "step": 6261
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 9.17868472262949e-06,
+      "loss": 0.0765,
+      "step": 6262
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 9.178551435502359e-06,
+      "loss": 0.1443,
+      "step": 6263
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 9.178418148375232e-06,
+      "loss": 0.1064,
+      "step": 6264
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 9.178284861248101e-06,
+      "loss": 0.105,
+      "step": 6265
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 9.178151574120972e-06,
+      "loss": 0.1183,
+      "step": 6266
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 9.178018286993843e-06,
+      "loss": 0.1246,
+      "step": 6267
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 9.177884999866714e-06,
+      "loss": 0.1013,
+      "step": 6268
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 9.177751712739585e-06,
+      "loss": 0.1172,
+      "step": 6269
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 9.177618425612456e-06,
+      "loss": 0.0982,
+      "step": 6270
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 9.177485138485325e-06,
+      "loss": 0.0908,
+      "step": 6271
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 9.177351851358196e-06,
+      "loss": 0.0757,
+      "step": 6272
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 9.177218564231067e-06,
+      "loss": 0.1048,
+      "step": 6273
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 9.177085277103938e-06,
+      "loss": 0.1022,
+      "step": 6274
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 9.17695198997681e-06,
+      "loss": 0.0964,
+      "step": 6275
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 9.176818702849678e-06,
+      "loss": 0.1092,
+      "step": 6276
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 9.176685415722551e-06,
+      "loss": 0.1122,
+      "step": 6277
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 9.17655212859542e-06,
+      "loss": 0.0934,
+      "step": 6278
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 9.176418841468291e-06,
+      "loss": 0.0837,
+      "step": 6279
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 9.176285554341162e-06,
+      "loss": 0.1429,
+      "step": 6280
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 9.176152267214033e-06,
+      "loss": 0.1185,
+      "step": 6281
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 9.176018980086904e-06,
+      "loss": 0.119,
+      "step": 6282
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 9.175885692959775e-06,
+      "loss": 0.1136,
+      "step": 6283
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 9.175752405832645e-06,
+      "loss": 0.098,
+      "step": 6284
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 9.175619118705516e-06,
+      "loss": 0.1294,
+      "step": 6285
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 9.175485831578387e-06,
+      "loss": 0.0882,
+      "step": 6286
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 9.175352544451258e-06,
+      "loss": 0.11,
+      "step": 6287
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 9.175219257324129e-06,
+      "loss": 0.0894,
+      "step": 6288
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 9.175085970197e-06,
+      "loss": 0.1443,
+      "step": 6289
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 9.174952683069869e-06,
+      "loss": 0.1012,
+      "step": 6290
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 9.17481939594274e-06,
+      "loss": 0.1289,
+      "step": 6291
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 9.174686108815611e-06,
+      "loss": 0.1544,
+      "step": 6292
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 9.174552821688482e-06,
+      "loss": 0.1184,
+      "step": 6293
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 9.174419534561353e-06,
+      "loss": 0.1209,
+      "step": 6294
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 9.174286247434222e-06,
+      "loss": 0.0881,
+      "step": 6295
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 9.174152960307095e-06,
+      "loss": 0.1052,
+      "step": 6296
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 9.174019673179964e-06,
+      "loss": 0.0834,
+      "step": 6297
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 9.173886386052835e-06,
+      "loss": 0.0879,
+      "step": 6298
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 9.173753098925706e-06,
+      "loss": 0.0957,
+      "step": 6299
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 9.173619811798577e-06,
+      "loss": 0.089,
+      "step": 6300
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 9.173486524671448e-06,
+      "loss": 0.1027,
+      "step": 6301
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 9.17335323754432e-06,
+      "loss": 0.117,
+      "step": 6302
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 9.173219950417189e-06,
+      "loss": 0.0811,
+      "step": 6303
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 9.173086663290061e-06,
+      "loss": 0.0952,
+      "step": 6304
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 9.17295337616293e-06,
+      "loss": 0.1176,
+      "step": 6305
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 9.172820089035802e-06,
+      "loss": 0.1097,
+      "step": 6306
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 9.172686801908673e-06,
+      "loss": 0.0864,
+      "step": 6307
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 9.172553514781544e-06,
+      "loss": 0.1478,
+      "step": 6308
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 9.172420227654415e-06,
+      "loss": 0.0789,
+      "step": 6309
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 9.172286940527284e-06,
+      "loss": 0.1067,
+      "step": 6310
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 9.172153653400155e-06,
+      "loss": 0.108,
+      "step": 6311
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 9.172020366273026e-06,
+      "loss": 0.1061,
+      "step": 6312
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 9.171887079145897e-06,
+      "loss": 0.082,
+      "step": 6313
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 9.171753792018768e-06,
+      "loss": 0.1323,
+      "step": 6314
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 9.171620504891639e-06,
+      "loss": 0.1066,
+      "step": 6315
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 9.171487217764508e-06,
+      "loss": 0.1245,
+      "step": 6316
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 9.17135393063738e-06,
+      "loss": 0.1028,
+      "step": 6317
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 9.17122064351025e-06,
+      "loss": 0.0999,
+      "step": 6318
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 9.171087356383121e-06,
+      "loss": 0.1147,
+      "step": 6319
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 9.170954069255992e-06,
+      "loss": 0.1164,
+      "step": 6320
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 9.170820782128863e-06,
+      "loss": 0.1185,
+      "step": 6321
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 9.170687495001734e-06,
+      "loss": 0.1156,
+      "step": 6322
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 9.170554207874605e-06,
+      "loss": 0.1383,
+      "step": 6323
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 9.170420920747474e-06,
+      "loss": 0.1627,
+      "step": 6324
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 9.170287633620345e-06,
+      "loss": 0.1141,
+      "step": 6325
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 9.170154346493216e-06,
+      "loss": 0.1188,
+      "step": 6326
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 9.170021059366087e-06,
+      "loss": 0.088,
+      "step": 6327
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 9.169887772238958e-06,
+      "loss": 0.1387,
+      "step": 6328
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 9.16975448511183e-06,
+      "loss": 0.0965,
+      "step": 6329
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 9.169621197984699e-06,
+      "loss": 0.1074,
+      "step": 6330
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 9.16948791085757e-06,
+      "loss": 0.1753,
+      "step": 6331
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 9.16935462373044e-06,
+      "loss": 0.124,
+      "step": 6332
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 9.169221336603312e-06,
+      "loss": 0.0761,
+      "step": 6333
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 9.169088049476183e-06,
+      "loss": 0.1199,
+      "step": 6334
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 9.168954762349052e-06,
+      "loss": 0.1193,
+      "step": 6335
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 9.168821475221925e-06,
+      "loss": 0.1415,
+      "step": 6336
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 9.168688188094794e-06,
+      "loss": 0.0753,
+      "step": 6337
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 9.168554900967665e-06,
+      "loss": 0.1234,
+      "step": 6338
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 9.168421613840536e-06,
+      "loss": 0.1015,
+      "step": 6339
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 9.168288326713407e-06,
+      "loss": 0.1321,
+      "step": 6340
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 9.168155039586278e-06,
+      "loss": 0.1411,
+      "step": 6341
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 9.168021752459149e-06,
+      "loss": 0.1255,
+      "step": 6342
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 9.167888465332018e-06,
+      "loss": 0.1079,
+      "step": 6343
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 9.167755178204891e-06,
+      "loss": 0.1033,
+      "step": 6344
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 9.16762189107776e-06,
+      "loss": 0.1643,
+      "step": 6345
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 9.167488603950631e-06,
+      "loss": 0.0942,
+      "step": 6346
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 9.167355316823502e-06,
+      "loss": 0.1112,
+      "step": 6347
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 9.167222029696373e-06,
+      "loss": 0.1257,
+      "step": 6348
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 9.167088742569244e-06,
+      "loss": 0.0744,
+      "step": 6349
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 9.166955455442113e-06,
+      "loss": 0.1399,
+      "step": 6350
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 9.166822168314984e-06,
+      "loss": 0.1099,
+      "step": 6351
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 9.166688881187855e-06,
+      "loss": 0.0999,
+      "step": 6352
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 9.166555594060726e-06,
+      "loss": 0.099,
+      "step": 6353
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 9.166422306933597e-06,
+      "loss": 0.0911,
+      "step": 6354
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 9.166289019806468e-06,
+      "loss": 0.1465,
+      "step": 6355
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 9.166155732679338e-06,
+      "loss": 0.0958,
+      "step": 6356
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 9.166022445552209e-06,
+      "loss": 0.1111,
+      "step": 6357
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 9.16588915842508e-06,
+      "loss": 0.0784,
+      "step": 6358
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 9.16575587129795e-06,
+      "loss": 0.1028,
+      "step": 6359
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 9.165622584170822e-06,
+      "loss": 0.1215,
+      "step": 6360
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 9.165489297043693e-06,
+      "loss": 0.0713,
+      "step": 6361
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 9.165356009916562e-06,
+      "loss": 0.0842,
+      "step": 6362
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 9.165222722789435e-06,
+      "loss": 0.1405,
+      "step": 6363
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 9.165089435662304e-06,
+      "loss": 0.0877,
+      "step": 6364
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 9.164956148535175e-06,
+      "loss": 0.1303,
+      "step": 6365
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 9.164822861408046e-06,
+      "loss": 0.1242,
+      "step": 6366
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 9.164689574280917e-06,
+      "loss": 0.0875,
+      "step": 6367
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 9.164556287153788e-06,
+      "loss": 0.0899,
+      "step": 6368
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 9.164423000026657e-06,
+      "loss": 0.1112,
+      "step": 6369
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 9.164289712899528e-06,
+      "loss": 0.1389,
+      "step": 6370
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 9.1641564257724e-06,
+      "loss": 0.1077,
+      "step": 6371
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 9.16402313864527e-06,
+      "loss": 0.0833,
+      "step": 6372
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 9.163889851518141e-06,
+      "loss": 0.15,
+      "step": 6373
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 9.163756564391012e-06,
+      "loss": 0.0923,
+      "step": 6374
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 9.163623277263882e-06,
+      "loss": 0.1173,
+      "step": 6375
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 9.163489990136754e-06,
+      "loss": 0.1146,
+      "step": 6376
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 9.163356703009624e-06,
+      "loss": 0.0894,
+      "step": 6377
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 9.163223415882495e-06,
+      "loss": 0.1439,
+      "step": 6378
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 9.163090128755366e-06,
+      "loss": 0.1167,
+      "step": 6379
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 9.162956841628237e-06,
+      "loss": 0.1122,
+      "step": 6380
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 9.162823554501108e-06,
+      "loss": 0.127,
+      "step": 6381
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 9.162690267373979e-06,
+      "loss": 0.1481,
+      "step": 6382
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 9.162556980246848e-06,
+      "loss": 0.1203,
+      "step": 6383
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 9.16242369311972e-06,
+      "loss": 0.0862,
+      "step": 6384
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 9.16229040599259e-06,
+      "loss": 0.1182,
+      "step": 6385
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 9.16215711886546e-06,
+      "loss": 0.0762,
+      "step": 6386
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 9.162023831738332e-06,
+      "loss": 0.1269,
+      "step": 6387
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 9.161890544611203e-06,
+      "loss": 0.1206,
+      "step": 6388
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 9.161757257484074e-06,
+      "loss": 0.1236,
+      "step": 6389
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 9.161623970356943e-06,
+      "loss": 0.1467,
+      "step": 6390
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 9.161490683229814e-06,
+      "loss": 0.1533,
+      "step": 6391
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 9.161357396102685e-06,
+      "loss": 0.1361,
+      "step": 6392
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 9.161224108975556e-06,
+      "loss": 0.0833,
+      "step": 6393
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 9.161090821848427e-06,
+      "loss": 0.108,
+      "step": 6394
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 9.160957534721298e-06,
+      "loss": 0.0849,
+      "step": 6395
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 9.160824247594167e-06,
+      "loss": 0.1131,
+      "step": 6396
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 9.160690960467038e-06,
+      "loss": 0.0949,
+      "step": 6397
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 9.16055767333991e-06,
+      "loss": 0.1494,
+      "step": 6398
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 9.16042438621278e-06,
+      "loss": 0.0877,
+      "step": 6399
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 9.160291099085651e-06,
+      "loss": 0.1011,
+      "step": 6400
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 9.160157811958522e-06,
+      "loss": 0.0968,
+      "step": 6401
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 9.160024524831392e-06,
+      "loss": 0.1107,
+      "step": 6402
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 9.159891237704264e-06,
+      "loss": 0.1043,
+      "step": 6403
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 9.159757950577134e-06,
+      "loss": 0.1629,
+      "step": 6404
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 9.159624663450005e-06,
+      "loss": 0.0619,
+      "step": 6405
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 9.159491376322876e-06,
+      "loss": 0.0848,
+      "step": 6406
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 9.159358089195747e-06,
+      "loss": 0.0705,
+      "step": 6407
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 9.159224802068618e-06,
+      "loss": 0.1107,
+      "step": 6408
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 9.159091514941487e-06,
+      "loss": 0.0958,
+      "step": 6409
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 9.158958227814358e-06,
+      "loss": 0.0959,
+      "step": 6410
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 9.158824940687229e-06,
+      "loss": 0.0972,
+      "step": 6411
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 9.1586916535601e-06,
+      "loss": 0.1144,
+      "step": 6412
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 9.158558366432971e-06,
+      "loss": 0.1583,
+      "step": 6413
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 9.158425079305842e-06,
+      "loss": 0.1192,
+      "step": 6414
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 9.158291792178711e-06,
+      "loss": 0.1539,
+      "step": 6415
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 9.158158505051584e-06,
+      "loss": 0.1012,
+      "step": 6416
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 9.158025217924453e-06,
+      "loss": 0.1109,
+      "step": 6417
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 9.157891930797324e-06,
+      "loss": 0.0901,
+      "step": 6418
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 9.157758643670195e-06,
+      "loss": 0.0861,
+      "step": 6419
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 9.157625356543066e-06,
+      "loss": 0.1009,
+      "step": 6420
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 9.157492069415937e-06,
+      "loss": 0.1358,
+      "step": 6421
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 9.157358782288808e-06,
+      "loss": 0.1699,
+      "step": 6422
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 9.157225495161677e-06,
+      "loss": 0.1079,
+      "step": 6423
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 9.157092208034548e-06,
+      "loss": 0.0868,
+      "step": 6424
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 9.15695892090742e-06,
+      "loss": 0.1197,
+      "step": 6425
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 9.15682563378029e-06,
+      "loss": 0.1028,
+      "step": 6426
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 9.156692346653161e-06,
+      "loss": 0.1218,
+      "step": 6427
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 9.15655905952603e-06,
+      "loss": 0.1046,
+      "step": 6428
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 9.156425772398902e-06,
+      "loss": 0.1114,
+      "step": 6429
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 9.156292485271773e-06,
+      "loss": 0.1124,
+      "step": 6430
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 9.156159198144644e-06,
+      "loss": 0.132,
+      "step": 6431
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 9.156025911017515e-06,
+      "loss": 0.1422,
+      "step": 6432
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 9.155892623890386e-06,
+      "loss": 0.0906,
+      "step": 6433
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 9.155759336763255e-06,
+      "loss": 0.1506,
+      "step": 6434
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 9.155626049636128e-06,
+      "loss": 0.1059,
+      "step": 6435
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 9.155492762508997e-06,
+      "loss": 0.1392,
+      "step": 6436
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 9.155359475381868e-06,
+      "loss": 0.1142,
+      "step": 6437
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 9.155226188254739e-06,
+      "loss": 0.1074,
+      "step": 6438
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 9.15509290112761e-06,
+      "loss": 0.1408,
+      "step": 6439
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 9.154959614000481e-06,
+      "loss": 0.0936,
+      "step": 6440
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 9.154826326873352e-06,
+      "loss": 0.0962,
+      "step": 6441
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 9.154693039746221e-06,
+      "loss": 0.1038,
+      "step": 6442
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 9.154559752619094e-06,
+      "loss": 0.1229,
+      "step": 6443
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 9.154426465491963e-06,
+      "loss": 0.0858,
+      "step": 6444
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 9.154293178364834e-06,
+      "loss": 0.1103,
+      "step": 6445
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 9.154159891237705e-06,
+      "loss": 0.1041,
+      "step": 6446
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 9.154026604110575e-06,
+      "loss": 0.0873,
+      "step": 6447
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 9.153893316983447e-06,
+      "loss": 0.0747,
+      "step": 6448
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 9.153760029856317e-06,
+      "loss": 0.0917,
+      "step": 6449
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 9.153626742729188e-06,
+      "loss": 0.1127,
+      "step": 6450
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 9.153493455602059e-06,
+      "loss": 0.1971,
+      "step": 6451
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 9.15336016847493e-06,
+      "loss": 0.0785,
+      "step": 6452
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 9.1532268813478e-06,
+      "loss": 0.0997,
+      "step": 6453
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 9.153093594220671e-06,
+      "loss": 0.1327,
+      "step": 6454
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 9.15296030709354e-06,
+      "loss": 0.154,
+      "step": 6455
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 9.152827019966413e-06,
+      "loss": 0.0985,
+      "step": 6456
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 9.152693732839283e-06,
+      "loss": 0.0907,
+      "step": 6457
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 9.152560445712154e-06,
+      "loss": 0.1068,
+      "step": 6458
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 9.152427158585025e-06,
+      "loss": 0.133,
+      "step": 6459
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 9.152293871457896e-06,
+      "loss": 0.1209,
+      "step": 6460
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 9.152160584330767e-06,
+      "loss": 0.0944,
+      "step": 6461
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 9.152027297203638e-06,
+      "loss": 0.1291,
+      "step": 6462
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 9.151894010076507e-06,
+      "loss": 0.0995,
+      "step": 6463
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 9.151760722949378e-06,
+      "loss": 0.1224,
+      "step": 6464
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 9.151627435822249e-06,
+      "loss": 0.1124,
+      "step": 6465
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 9.15149414869512e-06,
+      "loss": 0.1134,
+      "step": 6466
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 9.151360861567991e-06,
+      "loss": 0.1333,
+      "step": 6467
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 9.15122757444086e-06,
+      "loss": 0.1205,
+      "step": 6468
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 9.151094287313731e-06,
+      "loss": 0.0979,
+      "step": 6469
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 9.150961000186602e-06,
+      "loss": 0.1213,
+      "step": 6470
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 9.150827713059473e-06,
+      "loss": 0.1297,
+      "step": 6471
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 9.150694425932344e-06,
+      "loss": 0.0992,
+      "step": 6472
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 9.150561138805215e-06,
+      "loss": 0.105,
+      "step": 6473
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 9.150427851678085e-06,
+      "loss": 0.0794,
+      "step": 6474
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 9.150294564550957e-06,
+      "loss": 0.0983,
+      "step": 6475
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 9.150161277423827e-06,
+      "loss": 0.0806,
+      "step": 6476
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 9.150027990296698e-06,
+      "loss": 0.1077,
+      "step": 6477
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 9.149894703169569e-06,
+      "loss": 0.1011,
+      "step": 6478
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 9.14976141604244e-06,
+      "loss": 0.135,
+      "step": 6479
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 9.14962812891531e-06,
+      "loss": 0.1203,
+      "step": 6480
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 9.149494841788182e-06,
+      "loss": 0.1128,
+      "step": 6481
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 9.149361554661051e-06,
+      "loss": 0.0827,
+      "step": 6482
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 9.149228267533924e-06,
+      "loss": 0.1301,
+      "step": 6483
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 9.149094980406793e-06,
+      "loss": 0.1059,
+      "step": 6484
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 9.148961693279664e-06,
+      "loss": 0.067,
+      "step": 6485
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 9.148828406152535e-06,
+      "loss": 0.1482,
+      "step": 6486
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 9.148695119025404e-06,
+      "loss": 0.1126,
+      "step": 6487
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 9.148561831898277e-06,
+      "loss": 0.0863,
+      "step": 6488
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 9.148428544771146e-06,
+      "loss": 0.0942,
+      "step": 6489
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 9.148295257644017e-06,
+      "loss": 0.1141,
+      "step": 6490
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 9.148161970516888e-06,
+      "loss": 0.1768,
+      "step": 6491
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 9.148028683389759e-06,
+      "loss": 0.1134,
+      "step": 6492
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 9.14789539626263e-06,
+      "loss": 0.1253,
+      "step": 6493
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 9.147762109135501e-06,
+      "loss": 0.1118,
+      "step": 6494
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 9.14762882200837e-06,
+      "loss": 0.1182,
+      "step": 6495
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 9.147495534881241e-06,
+      "loss": 0.1112,
+      "step": 6496
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 9.147362247754112e-06,
+      "loss": 0.1014,
+      "step": 6497
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 9.147228960626983e-06,
+      "loss": 0.1024,
+      "step": 6498
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 9.147095673499854e-06,
+      "loss": 0.0634,
+      "step": 6499
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 9.146962386372725e-06,
+      "loss": 0.1069,
+      "step": 6500
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 9.146829099245595e-06,
+      "loss": 0.1022,
+      "step": 6501
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 9.146695812118467e-06,
+      "loss": 0.0893,
+      "step": 6502
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 9.146562524991337e-06,
+      "loss": 0.0897,
+      "step": 6503
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 9.146429237864208e-06,
+      "loss": 0.0668,
+      "step": 6504
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 9.146295950737079e-06,
+      "loss": 0.1017,
+      "step": 6505
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 9.146162663609948e-06,
+      "loss": 0.0844,
+      "step": 6506
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 9.14602937648282e-06,
+      "loss": 0.1014,
+      "step": 6507
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 9.14589608935569e-06,
+      "loss": 0.0942,
+      "step": 6508
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 9.145762802228561e-06,
+      "loss": 0.1135,
+      "step": 6509
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 9.145629515101432e-06,
+      "loss": 0.0997,
+      "step": 6510
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 9.145496227974303e-06,
+      "loss": 0.1291,
+      "step": 6511
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 9.145362940847174e-06,
+      "loss": 0.1052,
+      "step": 6512
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 9.145229653720045e-06,
+      "loss": 0.1463,
+      "step": 6513
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 9.145096366592914e-06,
+      "loss": 0.1099,
+      "step": 6514
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 9.144963079465787e-06,
+      "loss": 0.1138,
+      "step": 6515
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 9.144829792338656e-06,
+      "loss": 0.0999,
+      "step": 6516
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 9.144696505211527e-06,
+      "loss": 0.0811,
+      "step": 6517
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 9.144563218084398e-06,
+      "loss": 0.1226,
+      "step": 6518
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 9.14442993095727e-06,
+      "loss": 0.095,
+      "step": 6519
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 9.14429664383014e-06,
+      "loss": 0.0867,
+      "step": 6520
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 9.144163356703011e-06,
+      "loss": 0.0851,
+      "step": 6521
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 9.14403006957588e-06,
+      "loss": 0.1058,
+      "step": 6522
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 9.143896782448753e-06,
+      "loss": 0.1135,
+      "step": 6523
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 9.143763495321622e-06,
+      "loss": 0.0722,
+      "step": 6524
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 9.143630208194493e-06,
+      "loss": 0.1193,
+      "step": 6525
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 9.143496921067364e-06,
+      "loss": 0.2118,
+      "step": 6526
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 9.143363633940234e-06,
+      "loss": 0.1119,
+      "step": 6527
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 9.143230346813105e-06,
+      "loss": 0.1335,
+      "step": 6528
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 9.143097059685976e-06,
+      "loss": 0.1647,
+      "step": 6529
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 9.142963772558847e-06,
+      "loss": 0.1231,
+      "step": 6530
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 9.142830485431718e-06,
+      "loss": 0.1032,
+      "step": 6531
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 9.142697198304589e-06,
+      "loss": 0.1288,
+      "step": 6532
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 9.142563911177458e-06,
+      "loss": 0.0978,
+      "step": 6533
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 9.14243062405033e-06,
+      "loss": 0.1235,
+      "step": 6534
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 9.1422973369232e-06,
+      "loss": 0.1115,
+      "step": 6535
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 9.142164049796071e-06,
+      "loss": 0.0955,
+      "step": 6536
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 9.142030762668942e-06,
+      "loss": 0.1201,
+      "step": 6537
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 9.141897475541813e-06,
+      "loss": 0.088,
+      "step": 6538
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 9.141764188414684e-06,
+      "loss": 0.1209,
+      "step": 6539
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 9.141630901287555e-06,
+      "loss": 0.0958,
+      "step": 6540
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 9.141497614160424e-06,
+      "loss": 0.132,
+      "step": 6541
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 9.141364327033297e-06,
+      "loss": 0.0886,
+      "step": 6542
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 9.141231039906166e-06,
+      "loss": 0.1268,
+      "step": 6543
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 9.141097752779037e-06,
+      "loss": 0.1482,
+      "step": 6544
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 9.140964465651908e-06,
+      "loss": 0.1,
+      "step": 6545
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 9.140831178524778e-06,
+      "loss": 0.1674,
+      "step": 6546
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 9.14069789139765e-06,
+      "loss": 0.0956,
+      "step": 6547
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 9.14056460427052e-06,
+      "loss": 0.1588,
+      "step": 6548
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 9.14043131714339e-06,
+      "loss": 0.1069,
+      "step": 6549
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 9.140298030016262e-06,
+      "loss": 0.086,
+      "step": 6550
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 9.140164742889133e-06,
+      "loss": 0.1375,
+      "step": 6551
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 9.140031455762004e-06,
+      "loss": 0.1006,
+      "step": 6552
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 9.139898168634875e-06,
+      "loss": 0.1216,
+      "step": 6553
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 9.139764881507744e-06,
+      "loss": 0.065,
+      "step": 6554
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 9.139631594380617e-06,
+      "loss": 0.101,
+      "step": 6555
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 9.139498307253486e-06,
+      "loss": 0.0836,
+      "step": 6556
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 9.139365020126357e-06,
+      "loss": 0.1379,
+      "step": 6557
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 9.139231732999228e-06,
+      "loss": 0.128,
+      "step": 6558
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 9.139098445872099e-06,
+      "loss": 0.107,
+      "step": 6559
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 9.13896515874497e-06,
+      "loss": 0.1052,
+      "step": 6560
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 9.13883187161784e-06,
+      "loss": 0.1125,
+      "step": 6561
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 9.13869858449071e-06,
+      "loss": 0.0964,
+      "step": 6562
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 9.138565297363581e-06,
+      "loss": 0.1479,
+      "step": 6563
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 9.138432010236452e-06,
+      "loss": 0.114,
+      "step": 6564
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 9.138298723109323e-06,
+      "loss": 0.1334,
+      "step": 6565
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 9.138165435982194e-06,
+      "loss": 0.0809,
+      "step": 6566
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 9.138032148855063e-06,
+      "loss": 0.1364,
+      "step": 6567
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 9.137898861727934e-06,
+      "loss": 0.1718,
+      "step": 6568
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 9.137765574600805e-06,
+      "loss": 0.0892,
+      "step": 6569
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 9.137632287473676e-06,
+      "loss": 0.12,
+      "step": 6570
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 9.137499000346547e-06,
+      "loss": 0.0875,
+      "step": 6571
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 9.137365713219418e-06,
+      "loss": 0.1251,
+      "step": 6572
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 9.137232426092288e-06,
+      "loss": 0.0752,
+      "step": 6573
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 9.13709913896516e-06,
+      "loss": 0.1082,
+      "step": 6574
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 9.13696585183803e-06,
+      "loss": 0.1239,
+      "step": 6575
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 9.1368325647109e-06,
+      "loss": 0.0802,
+      "step": 6576
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 9.136699277583772e-06,
+      "loss": 0.0885,
+      "step": 6577
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 9.136565990456643e-06,
+      "loss": 0.1052,
+      "step": 6578
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 9.136432703329514e-06,
+      "loss": 0.0919,
+      "step": 6579
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 9.136299416202385e-06,
+      "loss": 0.0933,
+      "step": 6580
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 9.136166129075254e-06,
+      "loss": 0.083,
+      "step": 6581
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 9.136032841948127e-06,
+      "loss": 0.0849,
+      "step": 6582
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 9.135899554820996e-06,
+      "loss": 0.1376,
+      "step": 6583
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 9.135766267693867e-06,
+      "loss": 0.1024,
+      "step": 6584
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 9.135632980566738e-06,
+      "loss": 0.1113,
+      "step": 6585
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 9.135499693439607e-06,
+      "loss": 0.0961,
+      "step": 6586
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 9.13536640631248e-06,
+      "loss": 0.1417,
+      "step": 6587
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 9.13523311918535e-06,
+      "loss": 0.1095,
+      "step": 6588
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 9.13509983205822e-06,
+      "loss": 0.1147,
+      "step": 6589
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 9.134966544931091e-06,
+      "loss": 0.1264,
+      "step": 6590
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 9.134833257803962e-06,
+      "loss": 0.0997,
+      "step": 6591
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 9.134699970676833e-06,
+      "loss": 0.095,
+      "step": 6592
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 9.134566683549704e-06,
+      "loss": 0.085,
+      "step": 6593
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 9.134433396422573e-06,
+      "loss": 0.1018,
+      "step": 6594
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 9.134300109295444e-06,
+      "loss": 0.0801,
+      "step": 6595
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 9.134166822168315e-06,
+      "loss": 0.0918,
+      "step": 6596
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 9.134033535041186e-06,
+      "loss": 0.1506,
+      "step": 6597
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 9.133900247914057e-06,
+      "loss": 0.0846,
+      "step": 6598
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 9.133766960786928e-06,
+      "loss": 0.1794,
+      "step": 6599
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 9.133633673659798e-06,
+      "loss": 0.1232,
+      "step": 6600
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 9.13350038653267e-06,
+      "loss": 0.1581,
+      "step": 6601
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 9.13336709940554e-06,
+      "loss": 0.1363,
+      "step": 6602
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 9.13323381227841e-06,
+      "loss": 0.1022,
+      "step": 6603
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 9.133100525151282e-06,
+      "loss": 0.0966,
+      "step": 6604
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 9.132967238024151e-06,
+      "loss": 0.0838,
+      "step": 6605
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 9.132833950897024e-06,
+      "loss": 0.0789,
+      "step": 6606
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 9.132700663769893e-06,
+      "loss": 0.0825,
+      "step": 6607
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 9.132567376642764e-06,
+      "loss": 0.1207,
+      "step": 6608
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 9.132434089515635e-06,
+      "loss": 0.0988,
+      "step": 6609
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 9.132300802388506e-06,
+      "loss": 0.1034,
+      "step": 6610
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 9.132167515261377e-06,
+      "loss": 0.1183,
+      "step": 6611
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 9.132034228134248e-06,
+      "loss": 0.1066,
+      "step": 6612
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 9.131900941007117e-06,
+      "loss": 0.0656,
+      "step": 6613
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 9.13176765387999e-06,
+      "loss": 0.0997,
+      "step": 6614
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 9.13163436675286e-06,
+      "loss": 0.1247,
+      "step": 6615
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 9.13150107962573e-06,
+      "loss": 0.0708,
+      "step": 6616
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 9.131367792498601e-06,
+      "loss": 0.0995,
+      "step": 6617
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 9.131234505371472e-06,
+      "loss": 0.1164,
+      "step": 6618
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 9.131101218244343e-06,
+      "loss": 0.1006,
+      "step": 6619
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 9.130967931117214e-06,
+      "loss": 0.1149,
+      "step": 6620
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 9.130834643990084e-06,
+      "loss": 0.1097,
+      "step": 6621
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 9.130701356862956e-06,
+      "loss": 0.1499,
+      "step": 6622
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 9.130568069735826e-06,
+      "loss": 0.0921,
+      "step": 6623
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 9.130434782608697e-06,
+      "loss": 0.1547,
+      "step": 6624
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 9.130301495481568e-06,
+      "loss": 0.1267,
+      "step": 6625
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 9.130168208354437e-06,
+      "loss": 0.0896,
+      "step": 6626
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 9.13003492122731e-06,
+      "loss": 0.1289,
+      "step": 6627
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 9.129901634100179e-06,
+      "loss": 0.1342,
+      "step": 6628
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 9.12976834697305e-06,
+      "loss": 0.1279,
+      "step": 6629
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 9.12963505984592e-06,
+      "loss": 0.1147,
+      "step": 6630
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 9.129501772718792e-06,
+      "loss": 0.1246,
+      "step": 6631
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 9.129368485591663e-06,
+      "loss": 0.08,
+      "step": 6632
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 9.129235198464534e-06,
+      "loss": 0.1252,
+      "step": 6633
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 9.129101911337403e-06,
+      "loss": 0.0834,
+      "step": 6634
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 9.128968624210274e-06,
+      "loss": 0.0938,
+      "step": 6635
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 9.128835337083145e-06,
+      "loss": 0.1104,
+      "step": 6636
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 9.128702049956016e-06,
+      "loss": 0.1397,
+      "step": 6637
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 9.128568762828887e-06,
+      "loss": 0.1085,
+      "step": 6638
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 9.128435475701758e-06,
+      "loss": 0.1326,
+      "step": 6639
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 9.128302188574627e-06,
+      "loss": 0.1206,
+      "step": 6640
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 9.1281689014475e-06,
+      "loss": 0.1003,
+      "step": 6641
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 9.12803561432037e-06,
+      "loss": 0.0982,
+      "step": 6642
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 9.12790232719324e-06,
+      "loss": 0.1224,
+      "step": 6643
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 9.127769040066111e-06,
+      "loss": 0.1244,
+      "step": 6644
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 9.12763575293898e-06,
+      "loss": 0.1077,
+      "step": 6645
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 9.127502465811853e-06,
+      "loss": 0.0899,
+      "step": 6646
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 9.127369178684723e-06,
+      "loss": 0.1273,
+      "step": 6647
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 9.127235891557594e-06,
+      "loss": 0.0838,
+      "step": 6648
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 9.127102604430465e-06,
+      "loss": 0.0614,
+      "step": 6649
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 9.126969317303336e-06,
+      "loss": 0.1256,
+      "step": 6650
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 9.126836030176207e-06,
+      "loss": 0.1217,
+      "step": 6651
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 9.126702743049078e-06,
+      "loss": 0.1037,
+      "step": 6652
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 9.126569455921947e-06,
+      "loss": 0.1166,
+      "step": 6653
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 9.12643616879482e-06,
+      "loss": 0.086,
+      "step": 6654
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 9.126302881667689e-06,
+      "loss": 0.1162,
+      "step": 6655
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 9.12616959454056e-06,
+      "loss": 0.1201,
+      "step": 6656
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 9.126036307413431e-06,
+      "loss": 0.1175,
+      "step": 6657
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 9.125903020286302e-06,
+      "loss": 0.1078,
+      "step": 6658
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 9.125769733159173e-06,
+      "loss": 0.1115,
+      "step": 6659
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 9.125636446032044e-06,
+      "loss": 0.0821,
+      "step": 6660
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 9.125503158904913e-06,
+      "loss": 0.0651,
+      "step": 6661
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 9.125369871777784e-06,
+      "loss": 0.114,
+      "step": 6662
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 9.125236584650655e-06,
+      "loss": 0.1168,
+      "step": 6663
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 9.125103297523526e-06,
+      "loss": 0.1108,
+      "step": 6664
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 9.124970010396397e-06,
+      "loss": 0.0863,
+      "step": 6665
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 9.124836723269266e-06,
+      "loss": 0.1087,
+      "step": 6666
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 9.124703436142137e-06,
+      "loss": 0.0876,
+      "step": 6667
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 9.124570149015008e-06,
+      "loss": 0.1201,
+      "step": 6668
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 9.12443686188788e-06,
+      "loss": 0.1232,
+      "step": 6669
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 9.12430357476075e-06,
+      "loss": 0.1368,
+      "step": 6670
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 9.124170287633621e-06,
+      "loss": 0.0786,
+      "step": 6671
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 9.12403700050649e-06,
+      "loss": 0.0995,
+      "step": 6672
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 9.123903713379363e-06,
+      "loss": 0.1358,
+      "step": 6673
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 9.123770426252233e-06,
+      "loss": 0.1219,
+      "step": 6674
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 9.123637139125104e-06,
+      "loss": 0.0968,
+      "step": 6675
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 9.123503851997975e-06,
+      "loss": 0.1128,
+      "step": 6676
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 9.123370564870846e-06,
+      "loss": 0.1585,
+      "step": 6677
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 9.123237277743717e-06,
+      "loss": 0.0782,
+      "step": 6678
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 9.123103990616588e-06,
+      "loss": 0.0857,
+      "step": 6679
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 9.122970703489457e-06,
+      "loss": 0.093,
+      "step": 6680
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 9.12283741636233e-06,
+      "loss": 0.1136,
+      "step": 6681
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 9.122704129235199e-06,
+      "loss": 0.1355,
+      "step": 6682
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 9.12257084210807e-06,
+      "loss": 0.0898,
+      "step": 6683
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 9.122437554980941e-06,
+      "loss": 0.0866,
+      "step": 6684
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 9.12230426785381e-06,
+      "loss": 0.1188,
+      "step": 6685
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 9.122170980726683e-06,
+      "loss": 0.0711,
+      "step": 6686
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 9.122037693599552e-06,
+      "loss": 0.1241,
+      "step": 6687
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 9.121904406472423e-06,
+      "loss": 0.1074,
+      "step": 6688
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 9.121771119345294e-06,
+      "loss": 0.0892,
+      "step": 6689
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 9.121637832218165e-06,
+      "loss": 0.1119,
+      "step": 6690
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 9.121504545091036e-06,
+      "loss": 0.1021,
+      "step": 6691
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 9.121371257963907e-06,
+      "loss": 0.0716,
+      "step": 6692
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 9.121237970836777e-06,
+      "loss": 0.0716,
+      "step": 6693
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 9.12110468370965e-06,
+      "loss": 0.1436,
+      "step": 6694
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 9.120971396582519e-06,
+      "loss": 0.1479,
+      "step": 6695
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 9.12083810945539e-06,
+      "loss": 0.1079,
+      "step": 6696
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 9.12070482232826e-06,
+      "loss": 0.0986,
+      "step": 6697
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 9.120571535201132e-06,
+      "loss": 0.1623,
+      "step": 6698
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 9.120438248074003e-06,
+      "loss": 0.1086,
+      "step": 6699
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 9.120304960946874e-06,
+      "loss": 0.1163,
+      "step": 6700
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 9.120171673819743e-06,
+      "loss": 0.1107,
+      "step": 6701
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 9.120038386692614e-06,
+      "loss": 0.1099,
+      "step": 6702
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 9.119905099565485e-06,
+      "loss": 0.0916,
+      "step": 6703
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 9.119771812438356e-06,
+      "loss": 0.0739,
+      "step": 6704
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 9.119638525311227e-06,
+      "loss": 0.0935,
+      "step": 6705
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 9.119505238184096e-06,
+      "loss": 0.104,
+      "step": 6706
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 9.119371951056967e-06,
+      "loss": 0.1213,
+      "step": 6707
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 9.119238663929838e-06,
+      "loss": 0.1306,
+      "step": 6708
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 9.119105376802709e-06,
+      "loss": 0.1047,
+      "step": 6709
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 9.11897208967558e-06,
+      "loss": 0.1196,
+      "step": 6710
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 9.118838802548451e-06,
+      "loss": 0.1055,
+      "step": 6711
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 9.11870551542132e-06,
+      "loss": 0.1224,
+      "step": 6712
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 9.118572228294193e-06,
+      "loss": 0.0956,
+      "step": 6713
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 9.118438941167062e-06,
+      "loss": 0.1409,
+      "step": 6714
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 9.118305654039933e-06,
+      "loss": 0.0883,
+      "step": 6715
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 9.118172366912804e-06,
+      "loss": 0.1305,
+      "step": 6716
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 9.118039079785675e-06,
+      "loss": 0.0891,
+      "step": 6717
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 9.117905792658546e-06,
+      "loss": 0.1395,
+      "step": 6718
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 9.117772505531417e-06,
+      "loss": 0.0913,
+      "step": 6719
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 9.117639218404287e-06,
+      "loss": 0.1048,
+      "step": 6720
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 9.11750593127716e-06,
+      "loss": 0.1234,
+      "step": 6721
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 9.117372644150029e-06,
+      "loss": 0.0914,
+      "step": 6722
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 9.1172393570229e-06,
+      "loss": 0.0848,
+      "step": 6723
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 9.11710606989577e-06,
+      "loss": 0.1298,
+      "step": 6724
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 9.11697278276864e-06,
+      "loss": 0.1208,
+      "step": 6725
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 9.116839495641513e-06,
+      "loss": 0.1021,
+      "step": 6726
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 9.116706208514382e-06,
+      "loss": 0.1056,
+      "step": 6727
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 9.116572921387253e-06,
+      "loss": 0.1098,
+      "step": 6728
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 9.116439634260124e-06,
+      "loss": 0.0975,
+      "step": 6729
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 9.116306347132995e-06,
+      "loss": 0.1206,
+      "step": 6730
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 9.116173060005866e-06,
+      "loss": 0.1137,
+      "step": 6731
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 9.116039772878737e-06,
+      "loss": 0.1211,
+      "step": 6732
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 9.115906485751606e-06,
+      "loss": 0.1183,
+      "step": 6733
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 9.115773198624477e-06,
+      "loss": 0.1958,
+      "step": 6734
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 9.115639911497348e-06,
+      "loss": 0.1388,
+      "step": 6735
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 9.11550662437022e-06,
+      "loss": 0.098,
+      "step": 6736
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 9.11537333724309e-06,
+      "loss": 0.0932,
+      "step": 6737
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 9.115240050115961e-06,
+      "loss": 0.1212,
+      "step": 6738
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 9.11510676298883e-06,
+      "loss": 0.0738,
+      "step": 6739
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 9.114973475861703e-06,
+      "loss": 0.119,
+      "step": 6740
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 9.114840188734572e-06,
+      "loss": 0.0922,
+      "step": 6741
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 9.114706901607443e-06,
+      "loss": 0.1271,
+      "step": 6742
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 9.114573614480314e-06,
+      "loss": 0.0991,
+      "step": 6743
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 9.114440327353184e-06,
+      "loss": 0.112,
+      "step": 6744
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 9.114307040226056e-06,
+      "loss": 0.1397,
+      "step": 6745
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 9.114173753098926e-06,
+      "loss": 0.1316,
+      "step": 6746
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 9.114040465971797e-06,
+      "loss": 0.1253,
+      "step": 6747
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 9.113907178844668e-06,
+      "loss": 0.1048,
+      "step": 6748
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 9.113773891717539e-06,
+      "loss": 0.1472,
+      "step": 6749
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 9.11364060459041e-06,
+      "loss": 0.0989,
+      "step": 6750
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 9.11350731746328e-06,
+      "loss": 0.0865,
+      "step": 6751
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 9.11337403033615e-06,
+      "loss": 0.0983,
+      "step": 6752
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 9.113240743209023e-06,
+      "loss": 0.0892,
+      "step": 6753
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 9.113107456081892e-06,
+      "loss": 0.124,
+      "step": 6754
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 9.112974168954763e-06,
+      "loss": 0.1019,
+      "step": 6755
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 9.112840881827634e-06,
+      "loss": 0.1353,
+      "step": 6756
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 9.112707594700505e-06,
+      "loss": 0.1125,
+      "step": 6757
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 9.112574307573376e-06,
+      "loss": 0.1141,
+      "step": 6758
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 9.112441020446247e-06,
+      "loss": 0.1023,
+      "step": 6759
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 9.112307733319116e-06,
+      "loss": 0.0838,
+      "step": 6760
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 9.112174446191987e-06,
+      "loss": 0.1086,
+      "step": 6761
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 9.112041159064858e-06,
+      "loss": 0.0975,
+      "step": 6762
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 9.11190787193773e-06,
+      "loss": 0.1231,
+      "step": 6763
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 9.1117745848106e-06,
+      "loss": 0.1101,
+      "step": 6764
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 9.11164129768347e-06,
+      "loss": 0.1275,
+      "step": 6765
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 9.111508010556342e-06,
+      "loss": 0.0904,
+      "step": 6766
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 9.111374723429212e-06,
+      "loss": 0.1418,
+      "step": 6767
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 9.111241436302083e-06,
+      "loss": 0.0702,
+      "step": 6768
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 9.111108149174954e-06,
+      "loss": 0.0796,
+      "step": 6769
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 9.110974862047825e-06,
+      "loss": 0.0844,
+      "step": 6770
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 9.110841574920695e-06,
+      "loss": 0.109,
+      "step": 6771
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 9.110708287793566e-06,
+      "loss": 0.0964,
+      "step": 6772
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 9.110575000666436e-06,
+      "loss": 0.13,
+      "step": 6773
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 9.110441713539307e-06,
+      "loss": 0.1175,
+      "step": 6774
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 9.110308426412178e-06,
+      "loss": 0.1022,
+      "step": 6775
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 9.110175139285049e-06,
+      "loss": 0.1303,
+      "step": 6776
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 9.11004185215792e-06,
+      "loss": 0.0776,
+      "step": 6777
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 9.10990856503079e-06,
+      "loss": 0.1347,
+      "step": 6778
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 9.10977527790366e-06,
+      "loss": 0.0694,
+      "step": 6779
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 9.109641990776531e-06,
+      "loss": 0.1222,
+      "step": 6780
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 9.109508703649402e-06,
+      "loss": 0.1656,
+      "step": 6781
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 9.109375416522273e-06,
+      "loss": 0.0948,
+      "step": 6782
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 9.109242129395144e-06,
+      "loss": 0.1501,
+      "step": 6783
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 9.109108842268013e-06,
+      "loss": 0.0943,
+      "step": 6784
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 9.108975555140886e-06,
+      "loss": 0.0867,
+      "step": 6785
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 9.108842268013755e-06,
+      "loss": 0.1047,
+      "step": 6786
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 9.108708980886626e-06,
+      "loss": 0.1004,
+      "step": 6787
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 9.108575693759497e-06,
+      "loss": 0.1004,
+      "step": 6788
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 9.108442406632368e-06,
+      "loss": 0.0803,
+      "step": 6789
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 9.10830911950524e-06,
+      "loss": 0.0924,
+      "step": 6790
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 9.10817583237811e-06,
+      "loss": 0.0926,
+      "step": 6791
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 9.10804254525098e-06,
+      "loss": 0.0922,
+      "step": 6792
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 9.107909258123852e-06,
+      "loss": 0.124,
+      "step": 6793
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 9.107775970996722e-06,
+      "loss": 0.0925,
+      "step": 6794
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 9.107642683869593e-06,
+      "loss": 0.1229,
+      "step": 6795
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 9.107509396742464e-06,
+      "loss": 0.1038,
+      "step": 6796
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 9.107376109615335e-06,
+      "loss": 0.1492,
+      "step": 6797
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 9.107242822488206e-06,
+      "loss": 0.1185,
+      "step": 6798
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 9.107109535361077e-06,
+      "loss": 0.0962,
+      "step": 6799
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 9.106976248233946e-06,
+      "loss": 0.0965,
+      "step": 6800
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 9.106842961106817e-06,
+      "loss": 0.0868,
+      "step": 6801
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 9.106709673979688e-06,
+      "loss": 0.1169,
+      "step": 6802
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 9.106576386852559e-06,
+      "loss": 0.1611,
+      "step": 6803
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 9.10644309972543e-06,
+      "loss": 0.1299,
+      "step": 6804
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 9.1063098125983e-06,
+      "loss": 0.1374,
+      "step": 6805
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 9.10617652547117e-06,
+      "loss": 0.083,
+      "step": 6806
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 9.106043238344041e-06,
+      "loss": 0.1309,
+      "step": 6807
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 9.105909951216912e-06,
+      "loss": 0.0902,
+      "step": 6808
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 9.105776664089783e-06,
+      "loss": 0.1337,
+      "step": 6809
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 9.105643376962654e-06,
+      "loss": 0.1126,
+      "step": 6810
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 9.105510089835523e-06,
+      "loss": 0.0761,
+      "step": 6811
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 9.105376802708396e-06,
+      "loss": 0.0819,
+      "step": 6812
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 9.105243515581265e-06,
+      "loss": 0.1369,
+      "step": 6813
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 9.105110228454136e-06,
+      "loss": 0.1243,
+      "step": 6814
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 9.104976941327007e-06,
+      "loss": 0.1489,
+      "step": 6815
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 9.104843654199878e-06,
+      "loss": 0.1417,
+      "step": 6816
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 9.10471036707275e-06,
+      "loss": 0.136,
+      "step": 6817
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 9.10457707994562e-06,
+      "loss": 0.1474,
+      "step": 6818
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 9.10444379281849e-06,
+      "loss": 0.1232,
+      "step": 6819
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 9.10431050569136e-06,
+      "loss": 0.1092,
+      "step": 6820
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 9.104177218564232e-06,
+      "loss": 0.1264,
+      "step": 6821
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 9.104043931437103e-06,
+      "loss": 0.1708,
+      "step": 6822
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 9.103910644309974e-06,
+      "loss": 0.1004,
+      "step": 6823
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 9.103777357182843e-06,
+      "loss": 0.1046,
+      "step": 6824
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 9.103644070055716e-06,
+      "loss": 0.1339,
+      "step": 6825
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 9.103510782928585e-06,
+      "loss": 0.1316,
+      "step": 6826
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 9.103377495801456e-06,
+      "loss": 0.0901,
+      "step": 6827
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 9.103244208674327e-06,
+      "loss": 0.0995,
+      "step": 6828
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 9.103110921547198e-06,
+      "loss": 0.1223,
+      "step": 6829
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 9.102977634420069e-06,
+      "loss": 0.0909,
+      "step": 6830
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 9.10284434729294e-06,
+      "loss": 0.1228,
+      "step": 6831
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 9.10271106016581e-06,
+      "loss": 0.0908,
+      "step": 6832
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 9.10257777303868e-06,
+      "loss": 0.1024,
+      "step": 6833
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 9.102444485911551e-06,
+      "loss": 0.0853,
+      "step": 6834
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 9.102311198784422e-06,
+      "loss": 0.109,
+      "step": 6835
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 9.102177911657293e-06,
+      "loss": 0.1363,
+      "step": 6836
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 9.102044624530164e-06,
+      "loss": 0.1039,
+      "step": 6837
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 9.101911337403034e-06,
+      "loss": 0.0851,
+      "step": 6838
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 9.101778050275905e-06,
+      "loss": 0.0715,
+      "step": 6839
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 9.101644763148775e-06,
+      "loss": 0.1061,
+      "step": 6840
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 9.101511476021646e-06,
+      "loss": 0.1126,
+      "step": 6841
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 9.101378188894517e-06,
+      "loss": 0.1062,
+      "step": 6842
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 9.101244901767387e-06,
+      "loss": 0.112,
+      "step": 6843
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 9.10111161464026e-06,
+      "loss": 0.1063,
+      "step": 6844
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 9.100978327513129e-06,
+      "loss": 0.1153,
+      "step": 6845
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 9.100845040386e-06,
+      "loss": 0.1231,
+      "step": 6846
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 9.10071175325887e-06,
+      "loss": 0.0791,
+      "step": 6847
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 9.100578466131742e-06,
+      "loss": 0.1375,
+      "step": 6848
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 9.100445179004613e-06,
+      "loss": 0.1042,
+      "step": 6849
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 9.100311891877484e-06,
+      "loss": 0.1352,
+      "step": 6850
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 9.100178604750353e-06,
+      "loss": 0.1164,
+      "step": 6851
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 9.100045317623226e-06,
+      "loss": 0.1092,
+      "step": 6852
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 9.099912030496095e-06,
+      "loss": 0.1214,
+      "step": 6853
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 9.099778743368966e-06,
+      "loss": 0.0817,
+      "step": 6854
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 9.099645456241837e-06,
+      "loss": 0.1034,
+      "step": 6855
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 9.099512169114708e-06,
+      "loss": 0.1484,
+      "step": 6856
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 9.099378881987579e-06,
+      "loss": 0.1322,
+      "step": 6857
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 9.09924559486045e-06,
+      "loss": 0.1293,
+      "step": 6858
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 9.09911230773332e-06,
+      "loss": 0.083,
+      "step": 6859
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 9.09897902060619e-06,
+      "loss": 0.1092,
+      "step": 6860
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 9.098845733479061e-06,
+      "loss": 0.0971,
+      "step": 6861
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 9.098712446351932e-06,
+      "loss": 0.1092,
+      "step": 6862
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 9.098579159224803e-06,
+      "loss": 0.0914,
+      "step": 6863
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 9.098445872097673e-06,
+      "loss": 0.099,
+      "step": 6864
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 9.098312584970545e-06,
+      "loss": 0.1327,
+      "step": 6865
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 9.098179297843415e-06,
+      "loss": 0.0716,
+      "step": 6866
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 9.098046010716286e-06,
+      "loss": 0.1274,
+      "step": 6867
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 9.097912723589157e-06,
+      "loss": 0.1181,
+      "step": 6868
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 9.097779436462028e-06,
+      "loss": 0.0936,
+      "step": 6869
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 9.097646149334899e-06,
+      "loss": 0.0942,
+      "step": 6870
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 9.09751286220777e-06,
+      "loss": 0.0927,
+      "step": 6871
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 9.097379575080639e-06,
+      "loss": 0.1262,
+      "step": 6872
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 9.09724628795351e-06,
+      "loss": 0.1413,
+      "step": 6873
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 9.097113000826381e-06,
+      "loss": 0.1287,
+      "step": 6874
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 9.096979713699252e-06,
+      "loss": 0.1338,
+      "step": 6875
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 9.096846426572123e-06,
+      "loss": 0.1338,
+      "step": 6876
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 9.096713139444994e-06,
+      "loss": 0.1214,
+      "step": 6877
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 9.096579852317863e-06,
+      "loss": 0.0828,
+      "step": 6878
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 9.096446565190734e-06,
+      "loss": 0.1196,
+      "step": 6879
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 9.096313278063605e-06,
+      "loss": 0.1063,
+      "step": 6880
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 9.096179990936476e-06,
+      "loss": 0.0835,
+      "step": 6881
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 9.096046703809347e-06,
+      "loss": 0.1188,
+      "step": 6882
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 9.095913416682216e-06,
+      "loss": 0.079,
+      "step": 6883
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 9.095780129555089e-06,
+      "loss": 0.1087,
+      "step": 6884
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 9.095646842427958e-06,
+      "loss": 0.1252,
+      "step": 6885
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 9.09551355530083e-06,
+      "loss": 0.0884,
+      "step": 6886
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 9.0953802681737e-06,
+      "loss": 0.1027,
+      "step": 6887
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 9.095246981046571e-06,
+      "loss": 0.1056,
+      "step": 6888
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 9.095113693919442e-06,
+      "loss": 0.1088,
+      "step": 6889
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 9.094980406792313e-06,
+      "loss": 0.0815,
+      "step": 6890
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 9.094847119665183e-06,
+      "loss": 0.1115,
+      "step": 6891
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 9.094713832538055e-06,
+      "loss": 0.0905,
+      "step": 6892
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 9.094580545410925e-06,
+      "loss": 0.1891,
+      "step": 6893
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 9.094447258283796e-06,
+      "loss": 0.0833,
+      "step": 6894
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 9.094313971156667e-06,
+      "loss": 0.1235,
+      "step": 6895
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 9.094180684029538e-06,
+      "loss": 0.099,
+      "step": 6896
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 9.094047396902409e-06,
+      "loss": 0.1161,
+      "step": 6897
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 9.093914109775278e-06,
+      "loss": 0.1036,
+      "step": 6898
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 9.093780822648149e-06,
+      "loss": 0.1037,
+      "step": 6899
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 9.09364753552102e-06,
+      "loss": 0.1018,
+      "step": 6900
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 9.093514248393891e-06,
+      "loss": 0.1094,
+      "step": 6901
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 9.093380961266762e-06,
+      "loss": 0.0944,
+      "step": 6902
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 9.093247674139633e-06,
+      "loss": 0.0846,
+      "step": 6903
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 9.093114387012502e-06,
+      "loss": 0.1363,
+      "step": 6904
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 9.092981099885373e-06,
+      "loss": 0.1602,
+      "step": 6905
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 9.092847812758244e-06,
+      "loss": 0.0879,
+      "step": 6906
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 9.092714525631115e-06,
+      "loss": 0.086,
+      "step": 6907
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 9.092581238503986e-06,
+      "loss": 0.12,
+      "step": 6908
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 9.092447951376857e-06,
+      "loss": 0.0883,
+      "step": 6909
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 9.092314664249726e-06,
+      "loss": 0.1393,
+      "step": 6910
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 9.0921813771226e-06,
+      "loss": 0.1443,
+      "step": 6911
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 9.092048089995468e-06,
+      "loss": 0.1185,
+      "step": 6912
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 9.09191480286834e-06,
+      "loss": 0.1005,
+      "step": 6913
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 9.09178151574121e-06,
+      "loss": 0.1055,
+      "step": 6914
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 9.091648228614081e-06,
+      "loss": 0.11,
+      "step": 6915
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 9.091514941486952e-06,
+      "loss": 0.1011,
+      "step": 6916
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 9.091381654359822e-06,
+      "loss": 0.1304,
+      "step": 6917
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 9.091248367232693e-06,
+      "loss": 0.09,
+      "step": 6918
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 9.091115080105564e-06,
+      "loss": 0.1118,
+      "step": 6919
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 9.090981792978435e-06,
+      "loss": 0.104,
+      "step": 6920
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 9.090848505851306e-06,
+      "loss": 0.1347,
+      "step": 6921
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 9.090715218724177e-06,
+      "loss": 0.1089,
+      "step": 6922
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 9.090581931597046e-06,
+      "loss": 0.1328,
+      "step": 6923
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 9.090448644469919e-06,
+      "loss": 0.1186,
+      "step": 6924
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 9.090315357342788e-06,
+      "loss": 0.1151,
+      "step": 6925
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 9.090182070215659e-06,
+      "loss": 0.1259,
+      "step": 6926
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 9.09004878308853e-06,
+      "loss": 0.099,
+      "step": 6927
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 9.089915495961401e-06,
+      "loss": 0.1015,
+      "step": 6928
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 9.089782208834272e-06,
+      "loss": 0.0889,
+      "step": 6929
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 9.089648921707143e-06,
+      "loss": 0.0735,
+      "step": 6930
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 9.089515634580012e-06,
+      "loss": 0.0776,
+      "step": 6931
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 9.089382347452885e-06,
+      "loss": 0.1054,
+      "step": 6932
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 9.089249060325754e-06,
+      "loss": 0.0932,
+      "step": 6933
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 9.089115773198625e-06,
+      "loss": 0.1179,
+      "step": 6934
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 9.088982486071496e-06,
+      "loss": 0.1151,
+      "step": 6935
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 9.088849198944367e-06,
+      "loss": 0.1117,
+      "step": 6936
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 9.088715911817238e-06,
+      "loss": 0.1177,
+      "step": 6937
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 9.088582624690108e-06,
+      "loss": 0.0986,
+      "step": 6938
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 9.088449337562979e-06,
+      "loss": 0.0951,
+      "step": 6939
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 9.08831605043585e-06,
+      "loss": 0.1213,
+      "step": 6940
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 9.08818276330872e-06,
+      "loss": 0.0866,
+      "step": 6941
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 9.088049476181592e-06,
+      "loss": 0.1288,
+      "step": 6942
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 9.087916189054463e-06,
+      "loss": 0.0852,
+      "step": 6943
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 9.087782901927332e-06,
+      "loss": 0.1081,
+      "step": 6944
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 9.087649614800203e-06,
+      "loss": 0.1024,
+      "step": 6945
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 9.087516327673074e-06,
+      "loss": 0.1009,
+      "step": 6946
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 9.087383040545945e-06,
+      "loss": 0.1081,
+      "step": 6947
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 9.087249753418816e-06,
+      "loss": 0.1132,
+      "step": 6948
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 9.087116466291687e-06,
+      "loss": 0.1434,
+      "step": 6949
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 9.086983179164556e-06,
+      "loss": 0.1136,
+      "step": 6950
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 9.086849892037429e-06,
+      "loss": 0.1056,
+      "step": 6951
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 9.086716604910298e-06,
+      "loss": 0.1092,
+      "step": 6952
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 9.086583317783169e-06,
+      "loss": 0.1356,
+      "step": 6953
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 9.08645003065604e-06,
+      "loss": 0.1295,
+      "step": 6954
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 9.086316743528911e-06,
+      "loss": 0.1076,
+      "step": 6955
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 9.086183456401782e-06,
+      "loss": 0.1133,
+      "step": 6956
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 9.086050169274651e-06,
+      "loss": 0.1088,
+      "step": 6957
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 9.085916882147522e-06,
+      "loss": 0.1299,
+      "step": 6958
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 9.085783595020393e-06,
+      "loss": 0.075,
+      "step": 6959
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 9.085650307893264e-06,
+      "loss": 0.1112,
+      "step": 6960
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 9.085517020766135e-06,
+      "loss": 0.1508,
+      "step": 6961
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 9.085383733639006e-06,
+      "loss": 0.1067,
+      "step": 6962
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 9.085250446511876e-06,
+      "loss": 0.1138,
+      "step": 6963
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 9.085117159384748e-06,
+      "loss": 0.14,
+      "step": 6964
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 9.084983872257618e-06,
+      "loss": 0.0959,
+      "step": 6965
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 9.084850585130489e-06,
+      "loss": 0.084,
+      "step": 6966
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 9.08471729800336e-06,
+      "loss": 0.1036,
+      "step": 6967
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 9.08458401087623e-06,
+      "loss": 0.1269,
+      "step": 6968
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 9.084450723749102e-06,
+      "loss": 0.1191,
+      "step": 6969
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 9.084317436621973e-06,
+      "loss": 0.1407,
+      "step": 6970
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 9.084184149494842e-06,
+      "loss": 0.1249,
+      "step": 6971
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 9.084050862367713e-06,
+      "loss": 0.0818,
+      "step": 6972
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 9.083917575240584e-06,
+      "loss": 0.1203,
+      "step": 6973
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 9.083784288113455e-06,
+      "loss": 0.1215,
+      "step": 6974
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 9.083651000986326e-06,
+      "loss": 0.1159,
+      "step": 6975
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 9.083517713859195e-06,
+      "loss": 0.0801,
+      "step": 6976
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 9.083384426732066e-06,
+      "loss": 0.145,
+      "step": 6977
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 9.083251139604937e-06,
+      "loss": 0.0915,
+      "step": 6978
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 9.083117852477808e-06,
+      "loss": 0.0945,
+      "step": 6979
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 9.08298456535068e-06,
+      "loss": 0.0691,
+      "step": 6980
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 9.08285127822355e-06,
+      "loss": 0.0698,
+      "step": 6981
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 9.08271799109642e-06,
+      "loss": 0.1108,
+      "step": 6982
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 9.082584703969292e-06,
+      "loss": 0.0857,
+      "step": 6983
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 9.082451416842161e-06,
+      "loss": 0.0999,
+      "step": 6984
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 9.082318129715032e-06,
+      "loss": 0.2028,
+      "step": 6985
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 9.082184842587903e-06,
+      "loss": 0.1196,
+      "step": 6986
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 9.082051555460774e-06,
+      "loss": 0.1076,
+      "step": 6987
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 9.081918268333645e-06,
+      "loss": 0.1406,
+      "step": 6988
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 9.081784981206516e-06,
+      "loss": 0.1167,
+      "step": 6989
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 9.081651694079386e-06,
+      "loss": 0.0759,
+      "step": 6990
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 9.081518406952258e-06,
+      "loss": 0.1155,
+      "step": 6991
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 9.081385119825128e-06,
+      "loss": 0.1523,
+      "step": 6992
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 9.081251832697999e-06,
+      "loss": 0.122,
+      "step": 6993
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 9.08111854557087e-06,
+      "loss": 0.1378,
+      "step": 6994
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 9.08098525844374e-06,
+      "loss": 0.0618,
+      "step": 6995
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 9.080851971316612e-06,
+      "loss": 0.1453,
+      "step": 6996
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 9.080718684189481e-06,
+      "loss": 0.1122,
+      "step": 6997
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 9.080585397062352e-06,
+      "loss": 0.1159,
+      "step": 6998
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 9.080452109935223e-06,
+      "loss": 0.0945,
+      "step": 6999
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 9.080318822808094e-06,
+      "loss": 0.0778,
+      "step": 7000
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 9.080185535680965e-06,
+      "loss": 0.1392,
+      "step": 7001
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 9.080052248553836e-06,
+      "loss": 0.1031,
+      "step": 7002
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 9.079918961426705e-06,
+      "loss": 0.1882,
+      "step": 7003
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 9.079785674299578e-06,
+      "loss": 0.1283,
+      "step": 7004
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 9.079652387172447e-06,
+      "loss": 0.0924,
+      "step": 7005
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 9.079519100045318e-06,
+      "loss": 0.1099,
+      "step": 7006
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 9.07938581291819e-06,
+      "loss": 0.0757,
+      "step": 7007
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 9.07925252579106e-06,
+      "loss": 0.1064,
+      "step": 7008
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 9.079119238663931e-06,
+      "loss": 0.0927,
+      "step": 7009
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 9.078985951536802e-06,
+      "loss": 0.0706,
+      "step": 7010
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 9.078852664409672e-06,
+      "loss": 0.0786,
+      "step": 7011
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 9.078719377282543e-06,
+      "loss": 0.0995,
+      "step": 7012
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 9.078586090155414e-06,
+      "loss": 0.1033,
+      "step": 7013
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 9.078452803028285e-06,
+      "loss": 0.0856,
+      "step": 7014
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 9.078319515901156e-06,
+      "loss": 0.1033,
+      "step": 7015
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 9.078186228774025e-06,
+      "loss": 0.2002,
+      "step": 7016
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 9.078052941646896e-06,
+      "loss": 0.1251,
+      "step": 7017
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 9.077919654519767e-06,
+      "loss": 0.1032,
+      "step": 7018
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 9.077786367392638e-06,
+      "loss": 0.1078,
+      "step": 7019
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 9.077653080265509e-06,
+      "loss": 0.1104,
+      "step": 7020
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 9.07751979313838e-06,
+      "loss": 0.1326,
+      "step": 7021
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 9.077386506011249e-06,
+      "loss": 0.1261,
+      "step": 7022
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 9.077253218884122e-06,
+      "loss": 0.0867,
+      "step": 7023
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 9.077119931756991e-06,
+      "loss": 0.1321,
+      "step": 7024
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 9.076986644629862e-06,
+      "loss": 0.1116,
+      "step": 7025
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 9.076853357502733e-06,
+      "loss": 0.1076,
+      "step": 7026
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 9.076720070375604e-06,
+      "loss": 0.0916,
+      "step": 7027
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 9.076586783248475e-06,
+      "loss": 0.0822,
+      "step": 7028
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 9.076453496121346e-06,
+      "loss": 0.0921,
+      "step": 7029
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 9.076320208994215e-06,
+      "loss": 0.1016,
+      "step": 7030
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 9.076186921867088e-06,
+      "loss": 0.118,
+      "step": 7031
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 9.076053634739957e-06,
+      "loss": 0.1199,
+      "step": 7032
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 9.075920347612828e-06,
+      "loss": 0.0817,
+      "step": 7033
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 9.0757870604857e-06,
+      "loss": 0.1454,
+      "step": 7034
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 9.075653773358569e-06,
+      "loss": 0.0931,
+      "step": 7035
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 9.075520486231441e-06,
+      "loss": 0.1102,
+      "step": 7036
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 9.07538719910431e-06,
+      "loss": 0.0876,
+      "step": 7037
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 9.075253911977182e-06,
+      "loss": 0.1161,
+      "step": 7038
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 9.075120624850053e-06,
+      "loss": 0.1136,
+      "step": 7039
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 9.074987337722924e-06,
+      "loss": 0.0861,
+      "step": 7040
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 9.074854050595795e-06,
+      "loss": 0.1119,
+      "step": 7041
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 9.074720763468666e-06,
+      "loss": 0.0954,
+      "step": 7042
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 9.074587476341535e-06,
+      "loss": 0.1173,
+      "step": 7043
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 9.074454189214406e-06,
+      "loss": 0.1048,
+      "step": 7044
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 9.074320902087277e-06,
+      "loss": 0.109,
+      "step": 7045
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 9.074187614960148e-06,
+      "loss": 0.0975,
+      "step": 7046
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 9.074054327833019e-06,
+      "loss": 0.0909,
+      "step": 7047
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 9.07392104070589e-06,
+      "loss": 0.1095,
+      "step": 7048
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 9.07378775357876e-06,
+      "loss": 0.1053,
+      "step": 7049
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 9.073654466451632e-06,
+      "loss": 0.1326,
+      "step": 7050
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 9.073521179324501e-06,
+      "loss": 0.1804,
+      "step": 7051
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 9.073387892197372e-06,
+      "loss": 0.0871,
+      "step": 7052
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 9.073254605070243e-06,
+      "loss": 0.0865,
+      "step": 7053
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 9.073121317943114e-06,
+      "loss": 0.1319,
+      "step": 7054
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 9.072988030815985e-06,
+      "loss": 0.0978,
+      "step": 7055
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 9.072854743688854e-06,
+      "loss": 0.1357,
+      "step": 7056
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 9.072721456561725e-06,
+      "loss": 0.093,
+      "step": 7057
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 9.072588169434596e-06,
+      "loss": 0.1072,
+      "step": 7058
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 9.072454882307467e-06,
+      "loss": 0.0986,
+      "step": 7059
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 9.072321595180338e-06,
+      "loss": 0.1069,
+      "step": 7060
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 9.07218830805321e-06,
+      "loss": 0.0847,
+      "step": 7061
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 9.072055020926079e-06,
+      "loss": 0.1113,
+      "step": 7062
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 9.071921733798951e-06,
+      "loss": 0.0951,
+      "step": 7063
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 9.07178844667182e-06,
+      "loss": 0.0848,
+      "step": 7064
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 9.071655159544692e-06,
+      "loss": 0.1002,
+      "step": 7065
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 9.071521872417563e-06,
+      "loss": 0.0775,
+      "step": 7066
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 9.071388585290434e-06,
+      "loss": 0.1057,
+      "step": 7067
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 9.071255298163305e-06,
+      "loss": 0.1014,
+      "step": 7068
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 9.071122011036176e-06,
+      "loss": 0.0825,
+      "step": 7069
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 9.070988723909045e-06,
+      "loss": 0.1199,
+      "step": 7070
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 9.070855436781916e-06,
+      "loss": 0.1559,
+      "step": 7071
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 9.070722149654787e-06,
+      "loss": 0.1026,
+      "step": 7072
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 9.070588862527658e-06,
+      "loss": 0.0956,
+      "step": 7073
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 9.070455575400529e-06,
+      "loss": 0.1153,
+      "step": 7074
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 9.070322288273398e-06,
+      "loss": 0.1265,
+      "step": 7075
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 9.07018900114627e-06,
+      "loss": 0.0873,
+      "step": 7076
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 9.07005571401914e-06,
+      "loss": 0.101,
+      "step": 7077
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 9.069922426892011e-06,
+      "loss": 0.1035,
+      "step": 7078
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 9.069789139764882e-06,
+      "loss": 0.1089,
+      "step": 7079
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 9.069655852637753e-06,
+      "loss": 0.1231,
+      "step": 7080
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 9.069522565510623e-06,
+      "loss": 0.0838,
+      "step": 7081
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 9.069389278383495e-06,
+      "loss": 0.0785,
+      "step": 7082
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 9.069255991256365e-06,
+      "loss": 0.1044,
+      "step": 7083
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 9.069122704129236e-06,
+      "loss": 0.1295,
+      "step": 7084
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 9.068989417002107e-06,
+      "loss": 0.1325,
+      "step": 7085
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 9.068856129874978e-06,
+      "loss": 0.1022,
+      "step": 7086
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 9.068722842747849e-06,
+      "loss": 0.1198,
+      "step": 7087
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 9.06858955562072e-06,
+      "loss": 0.1204,
+      "step": 7088
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 9.068456268493589e-06,
+      "loss": 0.0908,
+      "step": 7089
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 9.068322981366461e-06,
+      "loss": 0.1231,
+      "step": 7090
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 9.06818969423933e-06,
+      "loss": 0.1552,
+      "step": 7091
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 9.068056407112202e-06,
+      "loss": 0.1035,
+      "step": 7092
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 9.067923119985073e-06,
+      "loss": 0.1265,
+      "step": 7093
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 9.067789832857942e-06,
+      "loss": 0.1277,
+      "step": 7094
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 9.067656545730815e-06,
+      "loss": 0.1664,
+      "step": 7095
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 9.067523258603684e-06,
+      "loss": 0.0877,
+      "step": 7096
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 9.067389971476555e-06,
+      "loss": 0.1494,
+      "step": 7097
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 9.067256684349426e-06,
+      "loss": 0.1149,
+      "step": 7098
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 9.067123397222297e-06,
+      "loss": 0.0989,
+      "step": 7099
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 9.066990110095168e-06,
+      "loss": 0.1117,
+      "step": 7100
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 9.066856822968039e-06,
+      "loss": 0.1222,
+      "step": 7101
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 9.066723535840908e-06,
+      "loss": 0.0724,
+      "step": 7102
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 9.066590248713781e-06,
+      "loss": 0.1166,
+      "step": 7103
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 9.06645696158665e-06,
+      "loss": 0.0974,
+      "step": 7104
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 9.066323674459521e-06,
+      "loss": 0.1276,
+      "step": 7105
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 9.066190387332392e-06,
+      "loss": 0.1046,
+      "step": 7106
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 9.066057100205263e-06,
+      "loss": 0.1168,
+      "step": 7107
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 9.065923813078134e-06,
+      "loss": 0.0936,
+      "step": 7108
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 9.065790525951005e-06,
+      "loss": 0.1369,
+      "step": 7109
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 9.065657238823875e-06,
+      "loss": 0.0858,
+      "step": 7110
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 9.065523951696746e-06,
+      "loss": 0.0947,
+      "step": 7111
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 9.065390664569617e-06,
+      "loss": 0.0952,
+      "step": 7112
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 9.065257377442488e-06,
+      "loss": 0.099,
+      "step": 7113
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 9.065124090315359e-06,
+      "loss": 0.0913,
+      "step": 7114
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 9.064990803188228e-06,
+      "loss": 0.1299,
+      "step": 7115
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 9.064857516061099e-06,
+      "loss": 0.1064,
+      "step": 7116
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 9.06472422893397e-06,
+      "loss": 0.1207,
+      "step": 7117
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 9.064590941806841e-06,
+      "loss": 0.0852,
+      "step": 7118
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 9.064457654679712e-06,
+      "loss": 0.0991,
+      "step": 7119
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 9.064324367552583e-06,
+      "loss": 0.1154,
+      "step": 7120
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 9.064191080425452e-06,
+      "loss": 0.1033,
+      "step": 7121
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 9.064057793298325e-06,
+      "loss": 0.1251,
+      "step": 7122
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 9.063924506171194e-06,
+      "loss": 0.0867,
+      "step": 7123
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 9.063791219044065e-06,
+      "loss": 0.1034,
+      "step": 7124
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 9.063657931916936e-06,
+      "loss": 0.1118,
+      "step": 7125
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 9.063524644789807e-06,
+      "loss": 0.0942,
+      "step": 7126
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 9.063391357662678e-06,
+      "loss": 0.1474,
+      "step": 7127
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 9.063258070535549e-06,
+      "loss": 0.1135,
+      "step": 7128
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 9.063124783408418e-06,
+      "loss": 0.1182,
+      "step": 7129
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 9.062991496281291e-06,
+      "loss": 0.0871,
+      "step": 7130
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 9.06285820915416e-06,
+      "loss": 0.0911,
+      "step": 7131
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 9.062724922027031e-06,
+      "loss": 0.0679,
+      "step": 7132
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 9.062591634899902e-06,
+      "loss": 0.1034,
+      "step": 7133
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 9.062458347772772e-06,
+      "loss": 0.1283,
+      "step": 7134
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 9.062325060645644e-06,
+      "loss": 0.1414,
+      "step": 7135
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 9.062191773518514e-06,
+      "loss": 0.1197,
+      "step": 7136
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 9.062058486391385e-06,
+      "loss": 0.116,
+      "step": 7137
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 9.061925199264256e-06,
+      "loss": 0.1392,
+      "step": 7138
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 9.061791912137127e-06,
+      "loss": 0.1146,
+      "step": 7139
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 9.061658625009998e-06,
+      "loss": 0.0964,
+      "step": 7140
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 9.061525337882869e-06,
+      "loss": 0.1226,
+      "step": 7141
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 9.061392050755738e-06,
+      "loss": 0.0934,
+      "step": 7142
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 9.061258763628609e-06,
+      "loss": 0.1387,
+      "step": 7143
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 9.06112547650148e-06,
+      "loss": 0.1085,
+      "step": 7144
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 9.060992189374351e-06,
+      "loss": 0.0916,
+      "step": 7145
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 9.060858902247222e-06,
+      "loss": 0.1517,
+      "step": 7146
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 9.060725615120093e-06,
+      "loss": 0.1274,
+      "step": 7147
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 9.060592327992962e-06,
+      "loss": 0.1257,
+      "step": 7148
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 9.060459040865835e-06,
+      "loss": 0.1151,
+      "step": 7149
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 9.060325753738704e-06,
+      "loss": 0.0879,
+      "step": 7150
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 9.060192466611575e-06,
+      "loss": 0.1172,
+      "step": 7151
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 9.060059179484446e-06,
+      "loss": 0.1189,
+      "step": 7152
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 9.059925892357316e-06,
+      "loss": 0.1003,
+      "step": 7153
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 9.059792605230188e-06,
+      "loss": 0.1813,
+      "step": 7154
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 9.059659318103058e-06,
+      "loss": 0.1181,
+      "step": 7155
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 9.059526030975929e-06,
+      "loss": 0.1108,
+      "step": 7156
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 9.0593927438488e-06,
+      "loss": 0.0934,
+      "step": 7157
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 9.05925945672167e-06,
+      "loss": 0.1194,
+      "step": 7158
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 9.059126169594541e-06,
+      "loss": 0.0747,
+      "step": 7159
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 9.058992882467412e-06,
+      "loss": 0.1399,
+      "step": 7160
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 9.058859595340282e-06,
+      "loss": 0.0991,
+      "step": 7161
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 9.058726308213154e-06,
+      "loss": 0.0996,
+      "step": 7162
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 9.058593021086024e-06,
+      "loss": 0.1285,
+      "step": 7163
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 9.058459733958895e-06,
+      "loss": 0.1398,
+      "step": 7164
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 9.058326446831766e-06,
+      "loss": 0.1127,
+      "step": 7165
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 9.058193159704637e-06,
+      "loss": 0.0963,
+      "step": 7166
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 9.058059872577508e-06,
+      "loss": 0.0786,
+      "step": 7167
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 9.057926585450379e-06,
+      "loss": 0.1097,
+      "step": 7168
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 9.057793298323248e-06,
+      "loss": 0.1375,
+      "step": 7169
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 9.05766001119612e-06,
+      "loss": 0.084,
+      "step": 7170
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 9.05752672406899e-06,
+      "loss": 0.0728,
+      "step": 7171
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 9.057393436941861e-06,
+      "loss": 0.0954,
+      "step": 7172
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 9.057260149814732e-06,
+      "loss": 0.0948,
+      "step": 7173
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 9.057126862687601e-06,
+      "loss": 0.0958,
+      "step": 7174
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 9.056993575560474e-06,
+      "loss": 0.0978,
+      "step": 7175
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 9.056860288433343e-06,
+      "loss": 0.0778,
+      "step": 7176
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 9.056727001306214e-06,
+      "loss": 0.1182,
+      "step": 7177
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 9.056593714179085e-06,
+      "loss": 0.1077,
+      "step": 7178
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 9.056460427051956e-06,
+      "loss": 0.1165,
+      "step": 7179
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 9.056327139924827e-06,
+      "loss": 0.0963,
+      "step": 7180
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 9.056193852797698e-06,
+      "loss": 0.0954,
+      "step": 7181
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 9.056060565670568e-06,
+      "loss": 0.0957,
+      "step": 7182
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 9.055927278543439e-06,
+      "loss": 0.0964,
+      "step": 7183
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 9.05579399141631e-06,
+      "loss": 0.1013,
+      "step": 7184
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 9.05566070428918e-06,
+      "loss": 0.1111,
+      "step": 7185
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 9.055527417162052e-06,
+      "loss": 0.0838,
+      "step": 7186
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 9.055394130034923e-06,
+      "loss": 0.0704,
+      "step": 7187
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 9.055260842907792e-06,
+      "loss": 0.0922,
+      "step": 7188
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 9.055127555780665e-06,
+      "loss": 0.1034,
+      "step": 7189
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 9.054994268653534e-06,
+      "loss": 0.0971,
+      "step": 7190
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 9.054860981526405e-06,
+      "loss": 0.1006,
+      "step": 7191
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 9.054727694399276e-06,
+      "loss": 0.1064,
+      "step": 7192
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 9.054594407272145e-06,
+      "loss": 0.1213,
+      "step": 7193
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 9.054461120145018e-06,
+      "loss": 0.0874,
+      "step": 7194
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 9.054327833017887e-06,
+      "loss": 0.1073,
+      "step": 7195
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 9.054194545890758e-06,
+      "loss": 0.099,
+      "step": 7196
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 9.054061258763629e-06,
+      "loss": 0.1578,
+      "step": 7197
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 9.0539279716365e-06,
+      "loss": 0.1236,
+      "step": 7198
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 9.053794684509371e-06,
+      "loss": 0.0921,
+      "step": 7199
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 9.053661397382242e-06,
+      "loss": 0.0878,
+      "step": 7200
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 9.053528110255111e-06,
+      "loss": 0.1435,
+      "step": 7201
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 9.053394823127984e-06,
+      "loss": 0.0974,
+      "step": 7202
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 9.053261536000853e-06,
+      "loss": 0.0587,
+      "step": 7203
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 9.053128248873724e-06,
+      "loss": 0.1368,
+      "step": 7204
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 9.052994961746595e-06,
+      "loss": 0.0841,
+      "step": 7205
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 9.052861674619466e-06,
+      "loss": 0.1232,
+      "step": 7206
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 9.052728387492337e-06,
+      "loss": 0.1168,
+      "step": 7207
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 9.052595100365208e-06,
+      "loss": 0.0957,
+      "step": 7208
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 9.052461813238078e-06,
+      "loss": 0.0909,
+      "step": 7209
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 9.052328526110949e-06,
+      "loss": 0.0986,
+      "step": 7210
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 9.05219523898382e-06,
+      "loss": 0.0864,
+      "step": 7211
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 9.05206195185669e-06,
+      "loss": 0.1351,
+      "step": 7212
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 9.051928664729562e-06,
+      "loss": 0.1293,
+      "step": 7213
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 9.051795377602431e-06,
+      "loss": 0.1014,
+      "step": 7214
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 9.051662090475302e-06,
+      "loss": 0.0917,
+      "step": 7215
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 9.051528803348173e-06,
+      "loss": 0.0929,
+      "step": 7216
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 9.051395516221044e-06,
+      "loss": 0.0722,
+      "step": 7217
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 9.051262229093915e-06,
+      "loss": 0.0841,
+      "step": 7218
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 9.051128941966786e-06,
+      "loss": 0.1107,
+      "step": 7219
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 9.050995654839655e-06,
+      "loss": 0.1087,
+      "step": 7220
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 9.050862367712528e-06,
+      "loss": 0.1622,
+      "step": 7221
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 9.050729080585397e-06,
+      "loss": 0.1064,
+      "step": 7222
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 9.050595793458268e-06,
+      "loss": 0.0862,
+      "step": 7223
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 9.05046250633114e-06,
+      "loss": 0.1012,
+      "step": 7224
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 9.05032921920401e-06,
+      "loss": 0.1177,
+      "step": 7225
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 9.050195932076881e-06,
+      "loss": 0.0877,
+      "step": 7226
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 9.050062644949752e-06,
+      "loss": 0.1229,
+      "step": 7227
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 9.049929357822621e-06,
+      "loss": 0.0877,
+      "step": 7228
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 9.049796070695494e-06,
+      "loss": 0.1222,
+      "step": 7229
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 9.049662783568363e-06,
+      "loss": 0.1206,
+      "step": 7230
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 9.049529496441234e-06,
+      "loss": 0.1212,
+      "step": 7231
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 9.049396209314105e-06,
+      "loss": 0.0786,
+      "step": 7232
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 9.049262922186975e-06,
+      "loss": 0.0952,
+      "step": 7233
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 9.049129635059847e-06,
+      "loss": 0.1124,
+      "step": 7234
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 9.048996347932717e-06,
+      "loss": 0.0812,
+      "step": 7235
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 9.048863060805588e-06,
+      "loss": 0.0991,
+      "step": 7236
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 9.048729773678459e-06,
+      "loss": 0.1212,
+      "step": 7237
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 9.04859648655133e-06,
+      "loss": 0.1043,
+      "step": 7238
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 9.0484631994242e-06,
+      "loss": 0.1635,
+      "step": 7239
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 9.048329912297072e-06,
+      "loss": 0.0931,
+      "step": 7240
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 9.048196625169941e-06,
+      "loss": 0.1128,
+      "step": 7241
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 9.048063338042814e-06,
+      "loss": 0.14,
+      "step": 7242
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 9.047930050915683e-06,
+      "loss": 0.1331,
+      "step": 7243
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 9.047796763788554e-06,
+      "loss": 0.1185,
+      "step": 7244
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 9.047663476661425e-06,
+      "loss": 0.0805,
+      "step": 7245
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 9.047530189534296e-06,
+      "loss": 0.0829,
+      "step": 7246
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 9.047396902407167e-06,
+      "loss": 0.1082,
+      "step": 7247
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 9.047263615280038e-06,
+      "loss": 0.1122,
+      "step": 7248
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 9.047130328152907e-06,
+      "loss": 0.1093,
+      "step": 7249
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 9.046997041025778e-06,
+      "loss": 0.1208,
+      "step": 7250
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 9.04686375389865e-06,
+      "loss": 0.183,
+      "step": 7251
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 9.04673046677152e-06,
+      "loss": 0.1412,
+      "step": 7252
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 9.046597179644391e-06,
+      "loss": 0.0806,
+      "step": 7253
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 9.04646389251726e-06,
+      "loss": 0.1449,
+      "step": 7254
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 9.046330605390132e-06,
+      "loss": 0.1112,
+      "step": 7255
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 9.046197318263003e-06,
+      "loss": 0.0976,
+      "step": 7256
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 9.046064031135874e-06,
+      "loss": 0.0725,
+      "step": 7257
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 9.045930744008745e-06,
+      "loss": 0.1287,
+      "step": 7258
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 9.045797456881616e-06,
+      "loss": 0.1162,
+      "step": 7259
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 9.045664169754485e-06,
+      "loss": 0.1157,
+      "step": 7260
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 9.045530882627358e-06,
+      "loss": 0.1412,
+      "step": 7261
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 9.045397595500227e-06,
+      "loss": 0.1109,
+      "step": 7262
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 9.045264308373098e-06,
+      "loss": 0.0878,
+      "step": 7263
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 9.045131021245969e-06,
+      "loss": 0.1193,
+      "step": 7264
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 9.04499773411884e-06,
+      "loss": 0.1001,
+      "step": 7265
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 9.04486444699171e-06,
+      "loss": 0.131,
+      "step": 7266
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 9.044731159864582e-06,
+      "loss": 0.1077,
+      "step": 7267
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 9.044597872737451e-06,
+      "loss": 0.1038,
+      "step": 7268
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 9.044464585610324e-06,
+      "loss": 0.1052,
+      "step": 7269
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 9.044331298483193e-06,
+      "loss": 0.1065,
+      "step": 7270
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 9.044198011356064e-06,
+      "loss": 0.1014,
+      "step": 7271
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 9.044064724228935e-06,
+      "loss": 0.1107,
+      "step": 7272
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 9.043931437101804e-06,
+      "loss": 0.1316,
+      "step": 7273
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 9.043798149974677e-06,
+      "loss": 0.1287,
+      "step": 7274
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 9.043664862847546e-06,
+      "loss": 0.119,
+      "step": 7275
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 9.043531575720417e-06,
+      "loss": 0.1054,
+      "step": 7276
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 9.043398288593288e-06,
+      "loss": 0.099,
+      "step": 7277
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 9.04326500146616e-06,
+      "loss": 0.1165,
+      "step": 7278
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 9.04313171433903e-06,
+      "loss": 0.1227,
+      "step": 7279
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 9.042998427211901e-06,
+      "loss": 0.1031,
+      "step": 7280
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 9.04286514008477e-06,
+      "loss": 0.0948,
+      "step": 7281
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 9.042731852957642e-06,
+      "loss": 0.0694,
+      "step": 7282
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 9.042598565830513e-06,
+      "loss": 0.1022,
+      "step": 7283
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 9.042465278703384e-06,
+      "loss": 0.0959,
+      "step": 7284
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 9.042331991576255e-06,
+      "loss": 0.0962,
+      "step": 7285
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 9.042198704449126e-06,
+      "loss": 0.1217,
+      "step": 7286
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 9.042065417321995e-06,
+      "loss": 0.1209,
+      "step": 7287
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 9.041932130194868e-06,
+      "loss": 0.0825,
+      "step": 7288
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 9.041798843067737e-06,
+      "loss": 0.1055,
+      "step": 7289
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 9.041665555940608e-06,
+      "loss": 0.1374,
+      "step": 7290
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 9.041532268813479e-06,
+      "loss": 0.0937,
+      "step": 7291
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 9.041398981686348e-06,
+      "loss": 0.07,
+      "step": 7292
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 9.041265694559221e-06,
+      "loss": 0.1573,
+      "step": 7293
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 9.04113240743209e-06,
+      "loss": 0.0921,
+      "step": 7294
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 9.040999120304961e-06,
+      "loss": 0.104,
+      "step": 7295
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 9.040865833177832e-06,
+      "loss": 0.1172,
+      "step": 7296
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 9.040732546050703e-06,
+      "loss": 0.1254,
+      "step": 7297
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 9.040599258923574e-06,
+      "loss": 0.1214,
+      "step": 7298
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 9.040465971796445e-06,
+      "loss": 0.0815,
+      "step": 7299
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 9.040332684669314e-06,
+      "loss": 0.1371,
+      "step": 7300
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 9.040199397542187e-06,
+      "loss": 0.1238,
+      "step": 7301
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 9.040066110415056e-06,
+      "loss": 0.0976,
+      "step": 7302
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 9.039932823287927e-06,
+      "loss": 0.1152,
+      "step": 7303
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 9.039799536160798e-06,
+      "loss": 0.0967,
+      "step": 7304
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 9.03966624903367e-06,
+      "loss": 0.101,
+      "step": 7305
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 9.03953296190654e-06,
+      "loss": 0.1099,
+      "step": 7306
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 9.039399674779411e-06,
+      "loss": 0.0907,
+      "step": 7307
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 9.03926638765228e-06,
+      "loss": 0.147,
+      "step": 7308
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 9.039133100525152e-06,
+      "loss": 0.0999,
+      "step": 7309
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 9.038999813398023e-06,
+      "loss": 0.0878,
+      "step": 7310
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 9.038866526270894e-06,
+      "loss": 0.0949,
+      "step": 7311
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 9.038733239143765e-06,
+      "loss": 0.0826,
+      "step": 7312
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 9.038599952016634e-06,
+      "loss": 0.0847,
+      "step": 7313
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 9.038466664889505e-06,
+      "loss": 0.0783,
+      "step": 7314
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 9.038333377762376e-06,
+      "loss": 0.068,
+      "step": 7315
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 9.038200090635247e-06,
+      "loss": 0.1189,
+      "step": 7316
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 9.038066803508118e-06,
+      "loss": 0.063,
+      "step": 7317
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 9.037933516380989e-06,
+      "loss": 0.0973,
+      "step": 7318
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 9.037800229253858e-06,
+      "loss": 0.0992,
+      "step": 7319
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 9.037666942126731e-06,
+      "loss": 0.0878,
+      "step": 7320
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 9.0375336549996e-06,
+      "loss": 0.1582,
+      "step": 7321
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 9.037400367872471e-06,
+      "loss": 0.0831,
+      "step": 7322
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 9.037267080745342e-06,
+      "loss": 0.165,
+      "step": 7323
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 9.037133793618213e-06,
+      "loss": 0.1242,
+      "step": 7324
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 9.037000506491084e-06,
+      "loss": 0.117,
+      "step": 7325
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 9.036867219363955e-06,
+      "loss": 0.1013,
+      "step": 7326
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 9.036733932236825e-06,
+      "loss": 0.0963,
+      "step": 7327
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 9.036600645109697e-06,
+      "loss": 0.0905,
+      "step": 7328
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 9.036467357982567e-06,
+      "loss": 0.0725,
+      "step": 7329
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 9.036334070855438e-06,
+      "loss": 0.1047,
+      "step": 7330
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 9.036200783728309e-06,
+      "loss": 0.076,
+      "step": 7331
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 9.036067496601178e-06,
+      "loss": 0.0814,
+      "step": 7332
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 9.03593420947405e-06,
+      "loss": 0.132,
+      "step": 7333
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 9.03580092234692e-06,
+      "loss": 0.0949,
+      "step": 7334
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 9.03566763521979e-06,
+      "loss": 0.124,
+      "step": 7335
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 9.035534348092662e-06,
+      "loss": 0.1322,
+      "step": 7336
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 9.035401060965533e-06,
+      "loss": 0.0696,
+      "step": 7337
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 9.035267773838404e-06,
+      "loss": 0.1894,
+      "step": 7338
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 9.035134486711275e-06,
+      "loss": 0.1141,
+      "step": 7339
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 9.035001199584144e-06,
+      "loss": 0.0857,
+      "step": 7340
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 9.034867912457017e-06,
+      "loss": 0.078,
+      "step": 7341
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 9.034734625329886e-06,
+      "loss": 0.1005,
+      "step": 7342
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 9.034601338202757e-06,
+      "loss": 0.1383,
+      "step": 7343
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 9.034468051075628e-06,
+      "loss": 0.1002,
+      "step": 7344
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 9.034334763948499e-06,
+      "loss": 0.1053,
+      "step": 7345
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 9.03420147682137e-06,
+      "loss": 0.0884,
+      "step": 7346
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 9.034068189694241e-06,
+      "loss": 0.1094,
+      "step": 7347
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 9.03393490256711e-06,
+      "loss": 0.1012,
+      "step": 7348
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 9.033801615439981e-06,
+      "loss": 0.109,
+      "step": 7349
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 9.033668328312852e-06,
+      "loss": 0.0967,
+      "step": 7350
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 9.033535041185723e-06,
+      "loss": 0.1439,
+      "step": 7351
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 9.033401754058594e-06,
+      "loss": 0.0967,
+      "step": 7352
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 9.033268466931464e-06,
+      "loss": 0.0945,
+      "step": 7353
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 9.033135179804335e-06,
+      "loss": 0.0934,
+      "step": 7354
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 9.033001892677206e-06,
+      "loss": 0.1262,
+      "step": 7355
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 9.032868605550077e-06,
+      "loss": 0.1127,
+      "step": 7356
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 9.032735318422948e-06,
+      "loss": 0.1319,
+      "step": 7357
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 9.032602031295819e-06,
+      "loss": 0.1325,
+      "step": 7358
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 9.032468744168688e-06,
+      "loss": 0.0528,
+      "step": 7359
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 9.03233545704156e-06,
+      "loss": 0.086,
+      "step": 7360
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 9.03220216991443e-06,
+      "loss": 0.1001,
+      "step": 7361
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 9.032068882787301e-06,
+      "loss": 0.1148,
+      "step": 7362
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 9.031935595660172e-06,
+      "loss": 0.1196,
+      "step": 7363
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 9.031802308533043e-06,
+      "loss": 0.0985,
+      "step": 7364
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 9.031669021405914e-06,
+      "loss": 0.0925,
+      "step": 7365
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 9.031535734278785e-06,
+      "loss": 0.1039,
+      "step": 7366
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 9.031402447151654e-06,
+      "loss": 0.108,
+      "step": 7367
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 9.031269160024525e-06,
+      "loss": 0.1282,
+      "step": 7368
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 9.031135872897396e-06,
+      "loss": 0.1074,
+      "step": 7369
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 9.031002585770267e-06,
+      "loss": 0.1043,
+      "step": 7370
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 9.030869298643138e-06,
+      "loss": 0.1372,
+      "step": 7371
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 9.030736011516007e-06,
+      "loss": 0.1016,
+      "step": 7372
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 9.03060272438888e-06,
+      "loss": 0.1648,
+      "step": 7373
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 9.03046943726175e-06,
+      "loss": 0.077,
+      "step": 7374
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 9.03033615013462e-06,
+      "loss": 0.1286,
+      "step": 7375
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 9.030202863007491e-06,
+      "loss": 0.0853,
+      "step": 7376
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 9.030069575880362e-06,
+      "loss": 0.0903,
+      "step": 7377
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 9.029936288753233e-06,
+      "loss": 0.1103,
+      "step": 7378
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 9.029803001626104e-06,
+      "loss": 0.0882,
+      "step": 7379
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 9.029669714498974e-06,
+      "loss": 0.0981,
+      "step": 7380
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 9.029536427371845e-06,
+      "loss": 0.1308,
+      "step": 7381
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 9.029403140244716e-06,
+      "loss": 0.1004,
+      "step": 7382
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 9.029269853117587e-06,
+      "loss": 0.1276,
+      "step": 7383
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 9.029136565990458e-06,
+      "loss": 0.0945,
+      "step": 7384
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 9.029003278863329e-06,
+      "loss": 0.1209,
+      "step": 7385
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 9.028869991736198e-06,
+      "loss": 0.1018,
+      "step": 7386
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 9.02873670460907e-06,
+      "loss": 0.149,
+      "step": 7387
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 9.02860341748194e-06,
+      "loss": 0.0791,
+      "step": 7388
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 9.028470130354811e-06,
+      "loss": 0.1688,
+      "step": 7389
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 9.028336843227682e-06,
+      "loss": 0.1051,
+      "step": 7390
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 9.028203556100551e-06,
+      "loss": 0.1236,
+      "step": 7391
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 9.028070268973424e-06,
+      "loss": 0.0948,
+      "step": 7392
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 9.027936981846293e-06,
+      "loss": 0.1043,
+      "step": 7393
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 9.027803694719164e-06,
+      "loss": 0.1296,
+      "step": 7394
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 9.027670407592035e-06,
+      "loss": 0.0977,
+      "step": 7395
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 9.027537120464906e-06,
+      "loss": 0.0798,
+      "step": 7396
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 9.027403833337777e-06,
+      "loss": 0.0907,
+      "step": 7397
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 9.027270546210648e-06,
+      "loss": 0.0691,
+      "step": 7398
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 9.027137259083518e-06,
+      "loss": 0.1148,
+      "step": 7399
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 9.02700397195639e-06,
+      "loss": 0.0825,
+      "step": 7400
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 9.02687068482926e-06,
+      "loss": 0.1024,
+      "step": 7401
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 9.02673739770213e-06,
+      "loss": 0.1011,
+      "step": 7402
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 9.026604110575002e-06,
+      "loss": 0.1342,
+      "step": 7403
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 9.026470823447873e-06,
+      "loss": 0.0817,
+      "step": 7404
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 9.026337536320744e-06,
+      "loss": 0.0778,
+      "step": 7405
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 9.026204249193615e-06,
+      "loss": 0.1267,
+      "step": 7406
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 9.026070962066484e-06,
+      "loss": 0.0958,
+      "step": 7407
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 9.025937674939355e-06,
+      "loss": 0.1287,
+      "step": 7408
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 9.025804387812226e-06,
+      "loss": 0.0991,
+      "step": 7409
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 9.025671100685097e-06,
+      "loss": 0.1356,
+      "step": 7410
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 9.025537813557968e-06,
+      "loss": 0.0919,
+      "step": 7411
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 9.025404526430837e-06,
+      "loss": 0.1297,
+      "step": 7412
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 9.02527123930371e-06,
+      "loss": 0.1112,
+      "step": 7413
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 9.025137952176579e-06,
+      "loss": 0.1166,
+      "step": 7414
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 9.02500466504945e-06,
+      "loss": 0.0796,
+      "step": 7415
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 9.024871377922321e-06,
+      "loss": 0.0843,
+      "step": 7416
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 9.024738090795192e-06,
+      "loss": 0.0832,
+      "step": 7417
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 9.024604803668063e-06,
+      "loss": 0.1108,
+      "step": 7418
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 9.024471516540934e-06,
+      "loss": 0.0727,
+      "step": 7419
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 9.024338229413803e-06,
+      "loss": 0.092,
+      "step": 7420
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 9.024204942286674e-06,
+      "loss": 0.0654,
+      "step": 7421
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 9.024071655159545e-06,
+      "loss": 0.0995,
+      "step": 7422
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 9.023938368032416e-06,
+      "loss": 0.1194,
+      "step": 7423
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 9.023805080905287e-06,
+      "loss": 0.0958,
+      "step": 7424
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 9.023671793778158e-06,
+      "loss": 0.101,
+      "step": 7425
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 9.023538506651028e-06,
+      "loss": 0.0712,
+      "step": 7426
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 9.023405219523899e-06,
+      "loss": 0.0792,
+      "step": 7427
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 9.02327193239677e-06,
+      "loss": 0.1037,
+      "step": 7428
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 9.02313864526964e-06,
+      "loss": 0.118,
+      "step": 7429
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 9.023005358142512e-06,
+      "loss": 0.0656,
+      "step": 7430
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 9.022872071015381e-06,
+      "loss": 0.087,
+      "step": 7431
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 9.022738783888254e-06,
+      "loss": 0.1039,
+      "step": 7432
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 9.022605496761123e-06,
+      "loss": 0.0889,
+      "step": 7433
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 9.022472209633994e-06,
+      "loss": 0.0887,
+      "step": 7434
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 9.022338922506865e-06,
+      "loss": 0.1017,
+      "step": 7435
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 9.022205635379736e-06,
+      "loss": 0.1042,
+      "step": 7436
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 9.022072348252607e-06,
+      "loss": 0.0874,
+      "step": 7437
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 9.021939061125478e-06,
+      "loss": 0.0654,
+      "step": 7438
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 9.021805773998347e-06,
+      "loss": 0.0922,
+      "step": 7439
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 9.02167248687122e-06,
+      "loss": 0.1124,
+      "step": 7440
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 9.02153919974409e-06,
+      "loss": 0.1562,
+      "step": 7441
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 9.02140591261696e-06,
+      "loss": 0.0966,
+      "step": 7442
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 9.021272625489831e-06,
+      "loss": 0.084,
+      "step": 7443
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 9.021139338362702e-06,
+      "loss": 0.1153,
+      "step": 7444
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 9.021006051235573e-06,
+      "loss": 0.143,
+      "step": 7445
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 9.020872764108442e-06,
+      "loss": 0.0829,
+      "step": 7446
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 9.020739476981313e-06,
+      "loss": 0.1266,
+      "step": 7447
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 9.020606189854184e-06,
+      "loss": 0.0944,
+      "step": 7448
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 9.020472902727055e-06,
+      "loss": 0.1451,
+      "step": 7449
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 9.020339615599926e-06,
+      "loss": 0.1072,
+      "step": 7450
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 9.020206328472797e-06,
+      "loss": 0.1165,
+      "step": 7451
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 9.020073041345667e-06,
+      "loss": 0.128,
+      "step": 7452
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 9.019939754218538e-06,
+      "loss": 0.1086,
+      "step": 7453
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 9.019806467091409e-06,
+      "loss": 0.1311,
+      "step": 7454
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 9.01967317996428e-06,
+      "loss": 0.1169,
+      "step": 7455
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 9.01953989283715e-06,
+      "loss": 0.0846,
+      "step": 7456
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 9.019406605710022e-06,
+      "loss": 0.1165,
+      "step": 7457
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 9.019273318582891e-06,
+      "loss": 0.1187,
+      "step": 7458
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 9.019140031455764e-06,
+      "loss": 0.1055,
+      "step": 7459
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 9.019006744328633e-06,
+      "loss": 0.1067,
+      "step": 7460
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 9.018873457201504e-06,
+      "loss": 0.088,
+      "step": 7461
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 9.018740170074375e-06,
+      "loss": 0.1214,
+      "step": 7462
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 9.018606882947246e-06,
+      "loss": 0.0904,
+      "step": 7463
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 9.018473595820117e-06,
+      "loss": 0.1066,
+      "step": 7464
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 9.018340308692988e-06,
+      "loss": 0.1062,
+      "step": 7465
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 9.018207021565857e-06,
+      "loss": 0.1104,
+      "step": 7466
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 9.018073734438728e-06,
+      "loss": 0.1137,
+      "step": 7467
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 9.0179404473116e-06,
+      "loss": 0.1085,
+      "step": 7468
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 9.01780716018447e-06,
+      "loss": 0.1254,
+      "step": 7469
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 9.017673873057341e-06,
+      "loss": 0.12,
+      "step": 7470
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 9.01754058593021e-06,
+      "loss": 0.1228,
+      "step": 7471
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 9.017407298803083e-06,
+      "loss": 0.091,
+      "step": 7472
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 9.017274011675953e-06,
+      "loss": 0.1096,
+      "step": 7473
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 9.017140724548824e-06,
+      "loss": 0.1462,
+      "step": 7474
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 9.017007437421695e-06,
+      "loss": 0.1189,
+      "step": 7475
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 9.016874150294566e-06,
+      "loss": 0.1289,
+      "step": 7476
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 9.016740863167436e-06,
+      "loss": 0.1397,
+      "step": 7477
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 9.016607576040307e-06,
+      "loss": 0.0726,
+      "step": 7478
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 9.016474288913177e-06,
+      "loss": 0.1106,
+      "step": 7479
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 9.01634100178605e-06,
+      "loss": 0.1067,
+      "step": 7480
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 9.016207714658919e-06,
+      "loss": 0.113,
+      "step": 7481
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 9.01607442753179e-06,
+      "loss": 0.1072,
+      "step": 7482
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 9.01594114040466e-06,
+      "loss": 0.0914,
+      "step": 7483
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 9.015807853277532e-06,
+      "loss": 0.0673,
+      "step": 7484
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 9.015674566150403e-06,
+      "loss": 0.1198,
+      "step": 7485
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 9.015541279023272e-06,
+      "loss": 0.1104,
+      "step": 7486
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 9.015407991896143e-06,
+      "loss": 0.1018,
+      "step": 7487
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 9.015274704769014e-06,
+      "loss": 0.0984,
+      "step": 7488
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 9.015141417641885e-06,
+      "loss": 0.1026,
+      "step": 7489
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 9.015008130514756e-06,
+      "loss": 0.1512,
+      "step": 7490
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 9.014874843387627e-06,
+      "loss": 0.0877,
+      "step": 7491
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 9.014741556260496e-06,
+      "loss": 0.153,
+      "step": 7492
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 9.014608269133367e-06,
+      "loss": 0.1017,
+      "step": 7493
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 9.014474982006238e-06,
+      "loss": 0.1224,
+      "step": 7494
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 9.01434169487911e-06,
+      "loss": 0.1253,
+      "step": 7495
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 9.01420840775198e-06,
+      "loss": 0.1499,
+      "step": 7496
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 9.014075120624851e-06,
+      "loss": 0.0898,
+      "step": 7497
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 9.01394183349772e-06,
+      "loss": 0.1239,
+      "step": 7498
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 9.013808546370593e-06,
+      "loss": 0.0778,
+      "step": 7499
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 9.013675259243463e-06,
+      "loss": 0.0902,
+      "step": 7500
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 9.013541972116334e-06,
+      "loss": 0.0899,
+      "step": 7501
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 9.013408684989205e-06,
+      "loss": 0.1024,
+      "step": 7502
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 9.013275397862076e-06,
+      "loss": 0.1287,
+      "step": 7503
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 9.013142110734947e-06,
+      "loss": 0.1006,
+      "step": 7504
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 9.013008823607816e-06,
+      "loss": 0.0989,
+      "step": 7505
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 9.012875536480687e-06,
+      "loss": 0.1138,
+      "step": 7506
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 9.012742249353558e-06,
+      "loss": 0.1199,
+      "step": 7507
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 9.012608962226429e-06,
+      "loss": 0.0994,
+      "step": 7508
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 9.0124756750993e-06,
+      "loss": 0.0655,
+      "step": 7509
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 9.012342387972171e-06,
+      "loss": 0.091,
+      "step": 7510
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 9.01220910084504e-06,
+      "loss": 0.0972,
+      "step": 7511
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 9.012075813717913e-06,
+      "loss": 0.1149,
+      "step": 7512
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 9.011942526590782e-06,
+      "loss": 0.0836,
+      "step": 7513
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 9.011809239463653e-06,
+      "loss": 0.1321,
+      "step": 7514
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 9.011675952336524e-06,
+      "loss": 0.13,
+      "step": 7515
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 9.011542665209395e-06,
+      "loss": 0.1214,
+      "step": 7516
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 9.011409378082266e-06,
+      "loss": 0.0911,
+      "step": 7517
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 9.011276090955137e-06,
+      "loss": 0.1118,
+      "step": 7518
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 9.011142803828006e-06,
+      "loss": 0.0879,
+      "step": 7519
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 9.011009516700877e-06,
+      "loss": 0.1533,
+      "step": 7520
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 9.010876229573748e-06,
+      "loss": 0.0985,
+      "step": 7521
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 9.01074294244662e-06,
+      "loss": 0.0968,
+      "step": 7522
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 9.01060965531949e-06,
+      "loss": 0.1489,
+      "step": 7523
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 9.010476368192361e-06,
+      "loss": 0.1204,
+      "step": 7524
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 9.01034308106523e-06,
+      "loss": 0.1169,
+      "step": 7525
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 9.010209793938102e-06,
+      "loss": 0.0931,
+      "step": 7526
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 9.010076506810973e-06,
+      "loss": 0.1331,
+      "step": 7527
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 9.009943219683844e-06,
+      "loss": 0.0828,
+      "step": 7528
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 9.009809932556715e-06,
+      "loss": 0.1019,
+      "step": 7529
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 9.009676645429584e-06,
+      "loss": 0.0687,
+      "step": 7530
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 9.009543358302457e-06,
+      "loss": 0.1007,
+      "step": 7531
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 9.009410071175326e-06,
+      "loss": 0.1071,
+      "step": 7532
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 9.009276784048197e-06,
+      "loss": 0.1291,
+      "step": 7533
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 9.009143496921068e-06,
+      "loss": 0.1181,
+      "step": 7534
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 9.009010209793939e-06,
+      "loss": 0.0755,
+      "step": 7535
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 9.00887692266681e-06,
+      "loss": 0.0848,
+      "step": 7536
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 9.008743635539681e-06,
+      "loss": 0.088,
+      "step": 7537
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 9.00861034841255e-06,
+      "loss": 0.0807,
+      "step": 7538
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 9.008477061285423e-06,
+      "loss": 0.0807,
+      "step": 7539
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 9.008343774158292e-06,
+      "loss": 0.1124,
+      "step": 7540
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 9.008210487031163e-06,
+      "loss": 0.0879,
+      "step": 7541
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 9.008077199904034e-06,
+      "loss": 0.1533,
+      "step": 7542
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 9.007943912776905e-06,
+      "loss": 0.0833,
+      "step": 7543
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 9.007810625649776e-06,
+      "loss": 0.1214,
+      "step": 7544
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 9.007677338522646e-06,
+      "loss": 0.0986,
+      "step": 7545
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 9.007544051395516e-06,
+      "loss": 0.1208,
+      "step": 7546
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 9.007410764268387e-06,
+      "loss": 0.1436,
+      "step": 7547
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 9.007277477141258e-06,
+      "loss": 0.0793,
+      "step": 7548
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 9.00714419001413e-06,
+      "loss": 0.0898,
+      "step": 7549
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 9.007010902887e-06,
+      "loss": 0.123,
+      "step": 7550
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 9.00687761575987e-06,
+      "loss": 0.1229,
+      "step": 7551
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 9.006744328632742e-06,
+      "loss": 0.0844,
+      "step": 7552
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 9.006611041505612e-06,
+      "loss": 0.1315,
+      "step": 7553
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 9.006477754378483e-06,
+      "loss": 0.0925,
+      "step": 7554
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 9.006344467251354e-06,
+      "loss": 0.1052,
+      "step": 7555
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 9.006211180124225e-06,
+      "loss": 0.0828,
+      "step": 7556
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 9.006077892997094e-06,
+      "loss": 0.1048,
+      "step": 7557
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 9.005944605869967e-06,
+      "loss": 0.0816,
+      "step": 7558
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 9.005811318742836e-06,
+      "loss": 0.0829,
+      "step": 7559
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 9.005678031615707e-06,
+      "loss": 0.0974,
+      "step": 7560
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 9.005544744488578e-06,
+      "loss": 0.1561,
+      "step": 7561
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 9.005411457361449e-06,
+      "loss": 0.1094,
+      "step": 7562
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 9.00527817023432e-06,
+      "loss": 0.1131,
+      "step": 7563
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 9.00514488310719e-06,
+      "loss": 0.1205,
+      "step": 7564
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 9.00501159598006e-06,
+      "loss": 0.0701,
+      "step": 7565
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 9.004878308852931e-06,
+      "loss": 0.1296,
+      "step": 7566
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 9.004745021725802e-06,
+      "loss": 0.0877,
+      "step": 7567
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 9.004611734598673e-06,
+      "loss": 0.0815,
+      "step": 7568
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 9.004478447471544e-06,
+      "loss": 0.11,
+      "step": 7569
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 9.004345160344414e-06,
+      "loss": 0.0681,
+      "step": 7570
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 9.004211873217286e-06,
+      "loss": 0.0691,
+      "step": 7571
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 9.004078586090156e-06,
+      "loss": 0.1249,
+      "step": 7572
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 9.003945298963027e-06,
+      "loss": 0.0865,
+      "step": 7573
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 9.003812011835898e-06,
+      "loss": 0.0988,
+      "step": 7574
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 9.003678724708769e-06,
+      "loss": 0.1093,
+      "step": 7575
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 9.00354543758164e-06,
+      "loss": 0.1457,
+      "step": 7576
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 9.00341215045451e-06,
+      "loss": 0.1262,
+      "step": 7577
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 9.00327886332738e-06,
+      "loss": 0.0833,
+      "step": 7578
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 9.003145576200253e-06,
+      "loss": 0.1138,
+      "step": 7579
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 9.003012289073122e-06,
+      "loss": 0.0941,
+      "step": 7580
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 9.002879001945993e-06,
+      "loss": 0.0881,
+      "step": 7581
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 9.002745714818864e-06,
+      "loss": 0.098,
+      "step": 7582
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 9.002612427691735e-06,
+      "loss": 0.1081,
+      "step": 7583
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 9.002479140564606e-06,
+      "loss": 0.0982,
+      "step": 7584
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 9.002345853437475e-06,
+      "loss": 0.1226,
+      "step": 7585
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 9.002212566310346e-06,
+      "loss": 0.1213,
+      "step": 7586
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 9.002079279183217e-06,
+      "loss": 0.1043,
+      "step": 7587
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 9.001945992056088e-06,
+      "loss": 0.103,
+      "step": 7588
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 9.001812704928959e-06,
+      "loss": 0.0847,
+      "step": 7589
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 9.00167941780183e-06,
+      "loss": 0.0865,
+      "step": 7590
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 9.0015461306747e-06,
+      "loss": 0.2048,
+      "step": 7591
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 9.00141284354757e-06,
+      "loss": 0.0986,
+      "step": 7592
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 9.001279556420441e-06,
+      "loss": 0.0882,
+      "step": 7593
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 9.001146269293312e-06,
+      "loss": 0.0894,
+      "step": 7594
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 9.001012982166183e-06,
+      "loss": 0.1292,
+      "step": 7595
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 9.000879695039054e-06,
+      "loss": 0.1113,
+      "step": 7596
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 9.000746407911924e-06,
+      "loss": 0.1136,
+      "step": 7597
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 9.000613120784796e-06,
+      "loss": 0.1106,
+      "step": 7598
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 9.000479833657666e-06,
+      "loss": 0.086,
+      "step": 7599
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 9.000346546530537e-06,
+      "loss": 0.1079,
+      "step": 7600
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 9.000213259403408e-06,
+      "loss": 0.1094,
+      "step": 7601
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 9.000079972276279e-06,
+      "loss": 0.1007,
+      "step": 7602
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.99994668514915e-06,
+      "loss": 0.096,
+      "step": 7603
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.999813398022019e-06,
+      "loss": 0.1043,
+      "step": 7604
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.99968011089489e-06,
+      "loss": 0.0917,
+      "step": 7605
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.999546823767761e-06,
+      "loss": 0.1457,
+      "step": 7606
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.999413536640632e-06,
+      "loss": 0.1048,
+      "step": 7607
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.999280249513503e-06,
+      "loss": 0.1082,
+      "step": 7608
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.999146962386374e-06,
+      "loss": 0.102,
+      "step": 7609
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.999013675259243e-06,
+      "loss": 0.0828,
+      "step": 7610
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.998880388132116e-06,
+      "loss": 0.1129,
+      "step": 7611
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.998747101004985e-06,
+      "loss": 0.1059,
+      "step": 7612
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.998613813877856e-06,
+      "loss": 0.1132,
+      "step": 7613
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.998480526750727e-06,
+      "loss": 0.1167,
+      "step": 7614
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.998347239623598e-06,
+      "loss": 0.0981,
+      "step": 7615
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.99821395249647e-06,
+      "loss": 0.1071,
+      "step": 7616
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.99808066536934e-06,
+      "loss": 0.1662,
+      "step": 7617
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.99794737824221e-06,
+      "loss": 0.1033,
+      "step": 7618
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.99781409111508e-06,
+      "loss": 0.1095,
+      "step": 7619
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.997680803987951e-06,
+      "loss": 0.1138,
+      "step": 7620
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.997547516860822e-06,
+      "loss": 0.0838,
+      "step": 7621
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.997414229733693e-06,
+      "loss": 0.0789,
+      "step": 7622
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.997280942606563e-06,
+      "loss": 0.1072,
+      "step": 7623
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.997147655479434e-06,
+      "loss": 0.108,
+      "step": 7624
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.997014368352305e-06,
+      "loss": 0.1182,
+      "step": 7625
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.996881081225176e-06,
+      "loss": 0.1353,
+      "step": 7626
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.996747794098047e-06,
+      "loss": 0.1147,
+      "step": 7627
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.996614506970918e-06,
+      "loss": 0.0915,
+      "step": 7628
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.996481219843787e-06,
+      "loss": 0.0841,
+      "step": 7629
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.99634793271666e-06,
+      "loss": 0.1188,
+      "step": 7630
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.996214645589529e-06,
+      "loss": 0.096,
+      "step": 7631
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.9960813584624e-06,
+      "loss": 0.0829,
+      "step": 7632
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.995948071335271e-06,
+      "loss": 0.0956,
+      "step": 7633
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.995814784208142e-06,
+      "loss": 0.0866,
+      "step": 7634
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.995681497081013e-06,
+      "loss": 0.0948,
+      "step": 7635
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.995548209953884e-06,
+      "loss": 0.1026,
+      "step": 7636
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.995414922826753e-06,
+      "loss": 0.0888,
+      "step": 7637
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.995281635699626e-06,
+      "loss": 0.1005,
+      "step": 7638
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.995148348572495e-06,
+      "loss": 0.1171,
+      "step": 7639
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.995015061445366e-06,
+      "loss": 0.116,
+      "step": 7640
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.994881774318237e-06,
+      "loss": 0.1018,
+      "step": 7641
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.994748487191107e-06,
+      "loss": 0.1154,
+      "step": 7642
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.99461520006398e-06,
+      "loss": 0.1381,
+      "step": 7643
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.994481912936849e-06,
+      "loss": 0.0899,
+      "step": 7644
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.99434862580972e-06,
+      "loss": 0.089,
+      "step": 7645
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.99421533868259e-06,
+      "loss": 0.1089,
+      "step": 7646
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.994082051555462e-06,
+      "loss": 0.1417,
+      "step": 7647
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.993948764428333e-06,
+      "loss": 0.1096,
+      "step": 7648
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.993815477301204e-06,
+      "loss": 0.1277,
+      "step": 7649
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.993682190174073e-06,
+      "loss": 0.0953,
+      "step": 7650
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.993548903046946e-06,
+      "loss": 0.111,
+      "step": 7651
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.993415615919815e-06,
+      "loss": 0.1168,
+      "step": 7652
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.993282328792686e-06,
+      "loss": 0.0854,
+      "step": 7653
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.993149041665557e-06,
+      "loss": 0.1041,
+      "step": 7654
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.993015754538428e-06,
+      "loss": 0.0849,
+      "step": 7655
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.992882467411299e-06,
+      "loss": 0.0927,
+      "step": 7656
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.99274918028417e-06,
+      "loss": 0.1158,
+      "step": 7657
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.992615893157039e-06,
+      "loss": 0.0983,
+      "step": 7658
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.99248260602991e-06,
+      "loss": 0.1004,
+      "step": 7659
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.992349318902781e-06,
+      "loss": 0.114,
+      "step": 7660
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.992216031775652e-06,
+      "loss": 0.1296,
+      "step": 7661
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.992082744648523e-06,
+      "loss": 0.1286,
+      "step": 7662
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.991949457521392e-06,
+      "loss": 0.1262,
+      "step": 7663
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.991816170394263e-06,
+      "loss": 0.0703,
+      "step": 7664
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.991682883267134e-06,
+      "loss": 0.1328,
+      "step": 7665
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.991549596140005e-06,
+      "loss": 0.1084,
+      "step": 7666
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.991416309012876e-06,
+      "loss": 0.1233,
+      "step": 7667
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.991283021885747e-06,
+      "loss": 0.1011,
+      "step": 7668
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.991149734758617e-06,
+      "loss": 0.0687,
+      "step": 7669
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.99101644763149e-06,
+      "loss": 0.1129,
+      "step": 7670
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.990883160504359e-06,
+      "loss": 0.085,
+      "step": 7671
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.99074987337723e-06,
+      "loss": 0.145,
+      "step": 7672
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.9906165862501e-06,
+      "loss": 0.0931,
+      "step": 7673
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.990483299122972e-06,
+      "loss": 0.0878,
+      "step": 7674
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.990350011995843e-06,
+      "loss": 0.0862,
+      "step": 7675
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.990216724868714e-06,
+      "loss": 0.0936,
+      "step": 7676
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.990083437741583e-06,
+      "loss": 0.1829,
+      "step": 7677
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.989950150614456e-06,
+      "loss": 0.089,
+      "step": 7678
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.989816863487325e-06,
+      "loss": 0.1121,
+      "step": 7679
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.989683576360196e-06,
+      "loss": 0.0981,
+      "step": 7680
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.989550289233067e-06,
+      "loss": 0.0814,
+      "step": 7681
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.989417002105936e-06,
+      "loss": 0.1742,
+      "step": 7682
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.989283714978809e-06,
+      "loss": 0.0746,
+      "step": 7683
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.989150427851678e-06,
+      "loss": 0.0885,
+      "step": 7684
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.98901714072455e-06,
+      "loss": 0.1195,
+      "step": 7685
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.98888385359742e-06,
+      "loss": 0.1625,
+      "step": 7686
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.988750566470291e-06,
+      "loss": 0.1571,
+      "step": 7687
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.988617279343162e-06,
+      "loss": 0.1369,
+      "step": 7688
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.988483992216033e-06,
+      "loss": 0.1183,
+      "step": 7689
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.988350705088902e-06,
+      "loss": 0.1447,
+      "step": 7690
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.988217417961773e-06,
+      "loss": 0.0943,
+      "step": 7691
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.988084130834644e-06,
+      "loss": 0.1034,
+      "step": 7692
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.987950843707515e-06,
+      "loss": 0.086,
+      "step": 7693
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.987817556580386e-06,
+      "loss": 0.0888,
+      "step": 7694
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.987684269453257e-06,
+      "loss": 0.1348,
+      "step": 7695
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.987550982326127e-06,
+      "loss": 0.1298,
+      "step": 7696
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.987417695199e-06,
+      "loss": 0.1333,
+      "step": 7697
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.987284408071869e-06,
+      "loss": 0.07,
+      "step": 7698
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.98715112094474e-06,
+      "loss": 0.1276,
+      "step": 7699
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.98701783381761e-06,
+      "loss": 0.1148,
+      "step": 7700
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.98688454669048e-06,
+      "loss": 0.0862,
+      "step": 7701
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.986751259563353e-06,
+      "loss": 0.1451,
+      "step": 7702
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.986617972436222e-06,
+      "loss": 0.1116,
+      "step": 7703
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.986484685309093e-06,
+      "loss": 0.1318,
+      "step": 7704
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.986351398181964e-06,
+      "loss": 0.1104,
+      "step": 7705
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.986218111054835e-06,
+      "loss": 0.1273,
+      "step": 7706
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.986084823927706e-06,
+      "loss": 0.141,
+      "step": 7707
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.985951536800577e-06,
+      "loss": 0.0862,
+      "step": 7708
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.985818249673446e-06,
+      "loss": 0.1184,
+      "step": 7709
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.985684962546319e-06,
+      "loss": 0.0854,
+      "step": 7710
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.985551675419188e-06,
+      "loss": 0.1066,
+      "step": 7711
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.98541838829206e-06,
+      "loss": 0.1064,
+      "step": 7712
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.98528510116493e-06,
+      "loss": 0.0982,
+      "step": 7713
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.985151814037801e-06,
+      "loss": 0.1641,
+      "step": 7714
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.985018526910672e-06,
+      "loss": 0.0896,
+      "step": 7715
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.984885239783543e-06,
+      "loss": 0.1118,
+      "step": 7716
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.984751952656413e-06,
+      "loss": 0.1075,
+      "step": 7717
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.984618665529285e-06,
+      "loss": 0.1215,
+      "step": 7718
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.984485378402155e-06,
+      "loss": 0.0857,
+      "step": 7719
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.984352091275026e-06,
+      "loss": 0.1315,
+      "step": 7720
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.984218804147897e-06,
+      "loss": 0.1141,
+      "step": 7721
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.984085517020766e-06,
+      "loss": 0.1011,
+      "step": 7722
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.983952229893639e-06,
+      "loss": 0.1117,
+      "step": 7723
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.983818942766508e-06,
+      "loss": 0.1022,
+      "step": 7724
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.983685655639379e-06,
+      "loss": 0.1002,
+      "step": 7725
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.98355236851225e-06,
+      "loss": 0.0739,
+      "step": 7726
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.98341908138512e-06,
+      "loss": 0.0994,
+      "step": 7727
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.983285794257992e-06,
+      "loss": 0.1258,
+      "step": 7728
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.983152507130863e-06,
+      "loss": 0.1472,
+      "step": 7729
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.983019220003732e-06,
+      "loss": 0.1324,
+      "step": 7730
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.982885932876603e-06,
+      "loss": 0.0807,
+      "step": 7731
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.982752645749474e-06,
+      "loss": 0.135,
+      "step": 7732
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.982619358622345e-06,
+      "loss": 0.1127,
+      "step": 7733
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.982486071495216e-06,
+      "loss": 0.122,
+      "step": 7734
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.982352784368087e-06,
+      "loss": 0.1294,
+      "step": 7735
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.982219497240956e-06,
+      "loss": 0.1049,
+      "step": 7736
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.982086210113829e-06,
+      "loss": 0.079,
+      "step": 7737
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.981952922986698e-06,
+      "loss": 0.0826,
+      "step": 7738
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.98181963585957e-06,
+      "loss": 0.1346,
+      "step": 7739
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.98168634873244e-06,
+      "loss": 0.1287,
+      "step": 7740
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.98155306160531e-06,
+      "loss": 0.1366,
+      "step": 7741
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.981419774478182e-06,
+      "loss": 0.1093,
+      "step": 7742
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.981286487351052e-06,
+      "loss": 0.0997,
+      "step": 7743
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.981153200223923e-06,
+      "loss": 0.1349,
+      "step": 7744
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.981019913096794e-06,
+      "loss": 0.1188,
+      "step": 7745
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.980886625969665e-06,
+      "loss": 0.113,
+      "step": 7746
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.980753338842536e-06,
+      "loss": 0.1066,
+      "step": 7747
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.980620051715407e-06,
+      "loss": 0.1102,
+      "step": 7748
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.980486764588276e-06,
+      "loss": 0.1267,
+      "step": 7749
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.980353477461149e-06,
+      "loss": 0.0915,
+      "step": 7750
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.980220190334018e-06,
+      "loss": 0.1178,
+      "step": 7751
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.980086903206889e-06,
+      "loss": 0.1065,
+      "step": 7752
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.97995361607976e-06,
+      "loss": 0.1087,
+      "step": 7753
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.979820328952631e-06,
+      "loss": 0.1082,
+      "step": 7754
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.979687041825502e-06,
+      "loss": 0.0948,
+      "step": 7755
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.979553754698373e-06,
+      "loss": 0.1028,
+      "step": 7756
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.979420467571242e-06,
+      "loss": 0.0985,
+      "step": 7757
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.979287180444113e-06,
+      "loss": 0.0921,
+      "step": 7758
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.979153893316984e-06,
+      "loss": 0.1035,
+      "step": 7759
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.979020606189855e-06,
+      "loss": 0.0956,
+      "step": 7760
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.978887319062726e-06,
+      "loss": 0.0919,
+      "step": 7761
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.978754031935595e-06,
+      "loss": 0.1359,
+      "step": 7762
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.978620744808466e-06,
+      "loss": 0.1048,
+      "step": 7763
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.978487457681337e-06,
+      "loss": 0.0888,
+      "step": 7764
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.978354170554208e-06,
+      "loss": 0.084,
+      "step": 7765
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.97822088342708e-06,
+      "loss": 0.1044,
+      "step": 7766
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.97808759629995e-06,
+      "loss": 0.1099,
+      "step": 7767
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.97795430917282e-06,
+      "loss": 0.0686,
+      "step": 7768
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.977821022045692e-06,
+      "loss": 0.1358,
+      "step": 7769
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.977687734918562e-06,
+      "loss": 0.1172,
+      "step": 7770
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.977554447791433e-06,
+      "loss": 0.1276,
+      "step": 7771
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.977421160664304e-06,
+      "loss": 0.1018,
+      "step": 7772
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.977287873537175e-06,
+      "loss": 0.1394,
+      "step": 7773
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.977154586410046e-06,
+      "loss": 0.0942,
+      "step": 7774
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.977021299282917e-06,
+      "loss": 0.089,
+      "step": 7775
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.976888012155786e-06,
+      "loss": 0.1728,
+      "step": 7776
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.976754725028659e-06,
+      "loss": 0.116,
+      "step": 7777
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.976621437901528e-06,
+      "loss": 0.1002,
+      "step": 7778
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.976488150774399e-06,
+      "loss": 0.1312,
+      "step": 7779
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.97635486364727e-06,
+      "loss": 0.129,
+      "step": 7780
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.97622157652014e-06,
+      "loss": 0.0846,
+      "step": 7781
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.976088289393012e-06,
+      "loss": 0.0867,
+      "step": 7782
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.975955002265881e-06,
+      "loss": 0.0959,
+      "step": 7783
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.975821715138752e-06,
+      "loss": 0.093,
+      "step": 7784
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.975688428011623e-06,
+      "loss": 0.1113,
+      "step": 7785
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.975555140884494e-06,
+      "loss": 0.0864,
+      "step": 7786
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.975421853757365e-06,
+      "loss": 0.1302,
+      "step": 7787
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.975288566630236e-06,
+      "loss": 0.0834,
+      "step": 7788
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.975155279503106e-06,
+      "loss": 0.1281,
+      "step": 7789
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.975021992375978e-06,
+      "loss": 0.0977,
+      "step": 7790
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.974888705248848e-06,
+      "loss": 0.0974,
+      "step": 7791
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.974755418121719e-06,
+      "loss": 0.1637,
+      "step": 7792
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.97462213099459e-06,
+      "loss": 0.1266,
+      "step": 7793
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.97448884386746e-06,
+      "loss": 0.0885,
+      "step": 7794
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.974355556740331e-06,
+      "loss": 0.1166,
+      "step": 7795
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.974222269613202e-06,
+      "loss": 0.164,
+      "step": 7796
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.974088982486072e-06,
+      "loss": 0.0868,
+      "step": 7797
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.973955695358943e-06,
+      "loss": 0.0992,
+      "step": 7798
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.973822408231814e-06,
+      "loss": 0.112,
+      "step": 7799
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.973689121104685e-06,
+      "loss": 0.0982,
+      "step": 7800
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.973555833977556e-06,
+      "loss": 0.0842,
+      "step": 7801
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.973422546850425e-06,
+      "loss": 0.1285,
+      "step": 7802
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.973289259723296e-06,
+      "loss": 0.1182,
+      "step": 7803
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.973155972596167e-06,
+      "loss": 0.1003,
+      "step": 7804
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.973022685469038e-06,
+      "loss": 0.0984,
+      "step": 7805
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.972889398341909e-06,
+      "loss": 0.0951,
+      "step": 7806
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.97275611121478e-06,
+      "loss": 0.0928,
+      "step": 7807
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.97262282408765e-06,
+      "loss": 0.1046,
+      "step": 7808
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.972489536960522e-06,
+      "loss": 0.0945,
+      "step": 7809
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.972356249833391e-06,
+      "loss": 0.1187,
+      "step": 7810
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.972222962706262e-06,
+      "loss": 0.1304,
+      "step": 7811
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.972089675579133e-06,
+      "loss": 0.1127,
+      "step": 7812
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.971956388452004e-06,
+      "loss": 0.1488,
+      "step": 7813
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.971823101324875e-06,
+      "loss": 0.072,
+      "step": 7814
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.971689814197746e-06,
+      "loss": 0.0838,
+      "step": 7815
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.971556527070616e-06,
+      "loss": 0.0997,
+      "step": 7816
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.971423239943488e-06,
+      "loss": 0.1052,
+      "step": 7817
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.971289952816358e-06,
+      "loss": 0.1265,
+      "step": 7818
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.971156665689229e-06,
+      "loss": 0.1186,
+      "step": 7819
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.9710233785621e-06,
+      "loss": 0.1118,
+      "step": 7820
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.970890091434969e-06,
+      "loss": 0.1312,
+      "step": 7821
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.970756804307842e-06,
+      "loss": 0.1015,
+      "step": 7822
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.970623517180711e-06,
+      "loss": 0.1327,
+      "step": 7823
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.970490230053582e-06,
+      "loss": 0.087,
+      "step": 7824
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.970356942926453e-06,
+      "loss": 0.1237,
+      "step": 7825
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.970223655799324e-06,
+      "loss": 0.1249,
+      "step": 7826
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.970090368672195e-06,
+      "loss": 0.0909,
+      "step": 7827
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.969957081545066e-06,
+      "loss": 0.0947,
+      "step": 7828
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.969823794417935e-06,
+      "loss": 0.0839,
+      "step": 7829
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.969690507290806e-06,
+      "loss": 0.0938,
+      "step": 7830
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.969557220163677e-06,
+      "loss": 0.1419,
+      "step": 7831
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.969423933036548e-06,
+      "loss": 0.0903,
+      "step": 7832
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.969290645909419e-06,
+      "loss": 0.1215,
+      "step": 7833
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.96915735878229e-06,
+      "loss": 0.1512,
+      "step": 7834
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.96902407165516e-06,
+      "loss": 0.1067,
+      "step": 7835
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.968890784528032e-06,
+      "loss": 0.1353,
+      "step": 7836
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.968757497400901e-06,
+      "loss": 0.1058,
+      "step": 7837
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.968624210273772e-06,
+      "loss": 0.0901,
+      "step": 7838
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.968490923146643e-06,
+      "loss": 0.1053,
+      "step": 7839
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.968357636019513e-06,
+      "loss": 0.0925,
+      "step": 7840
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.968224348892385e-06,
+      "loss": 0.1263,
+      "step": 7841
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.968091061765255e-06,
+      "loss": 0.1049,
+      "step": 7842
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.967957774638126e-06,
+      "loss": 0.1273,
+      "step": 7843
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.967824487510997e-06,
+      "loss": 0.1115,
+      "step": 7844
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.967691200383868e-06,
+      "loss": 0.1385,
+      "step": 7845
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.967557913256739e-06,
+      "loss": 0.1183,
+      "step": 7846
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.96742462612961e-06,
+      "loss": 0.1042,
+      "step": 7847
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.967291339002479e-06,
+      "loss": 0.0717,
+      "step": 7848
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.967158051875352e-06,
+      "loss": 0.1402,
+      "step": 7849
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.967024764748221e-06,
+      "loss": 0.1006,
+      "step": 7850
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.966891477621092e-06,
+      "loss": 0.1276,
+      "step": 7851
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.966758190493963e-06,
+      "loss": 0.0956,
+      "step": 7852
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.966624903366834e-06,
+      "loss": 0.1205,
+      "step": 7853
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.966491616239705e-06,
+      "loss": 0.1177,
+      "step": 7854
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.966358329112576e-06,
+      "loss": 0.0997,
+      "step": 7855
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.966225041985445e-06,
+      "loss": 0.1022,
+      "step": 7856
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.966091754858316e-06,
+      "loss": 0.1036,
+      "step": 7857
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.965958467731187e-06,
+      "loss": 0.0722,
+      "step": 7858
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.965825180604058e-06,
+      "loss": 0.1183,
+      "step": 7859
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.96569189347693e-06,
+      "loss": 0.0965,
+      "step": 7860
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.965558606349799e-06,
+      "loss": 0.1015,
+      "step": 7861
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.96542531922267e-06,
+      "loss": 0.0973,
+      "step": 7862
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.96529203209554e-06,
+      "loss": 0.1104,
+      "step": 7863
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.965158744968411e-06,
+      "loss": 0.1072,
+      "step": 7864
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.965025457841282e-06,
+      "loss": 0.1182,
+      "step": 7865
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.964892170714153e-06,
+      "loss": 0.0946,
+      "step": 7866
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.964758883587023e-06,
+      "loss": 0.096,
+      "step": 7867
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.964625596459895e-06,
+      "loss": 0.0739,
+      "step": 7868
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.964492309332765e-06,
+      "loss": 0.1178,
+      "step": 7869
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.964359022205636e-06,
+      "loss": 0.1107,
+      "step": 7870
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.964225735078507e-06,
+      "loss": 0.114,
+      "step": 7871
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.964092447951378e-06,
+      "loss": 0.1115,
+      "step": 7872
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.963959160824249e-06,
+      "loss": 0.107,
+      "step": 7873
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.96382587369712e-06,
+      "loss": 0.1294,
+      "step": 7874
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.963692586569989e-06,
+      "loss": 0.1225,
+      "step": 7875
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.963559299442862e-06,
+      "loss": 0.127,
+      "step": 7876
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.963426012315731e-06,
+      "loss": 0.1208,
+      "step": 7877
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.963292725188602e-06,
+      "loss": 0.1068,
+      "step": 7878
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.963159438061473e-06,
+      "loss": 0.0992,
+      "step": 7879
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.963026150934342e-06,
+      "loss": 0.1093,
+      "step": 7880
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.962892863807215e-06,
+      "loss": 0.0952,
+      "step": 7881
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.962759576680084e-06,
+      "loss": 0.1034,
+      "step": 7882
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.962626289552955e-06,
+      "loss": 0.0619,
+      "step": 7883
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.962493002425826e-06,
+      "loss": 0.1129,
+      "step": 7884
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.962359715298697e-06,
+      "loss": 0.1437,
+      "step": 7885
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.962226428171568e-06,
+      "loss": 0.1019,
+      "step": 7886
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.96209314104444e-06,
+      "loss": 0.1107,
+      "step": 7887
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.961959853917309e-06,
+      "loss": 0.0908,
+      "step": 7888
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 8.961826566790181e-06,
+      "loss": 0.0848,
+      "step": 7889
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 8.96169327966305e-06,
+      "loss": 0.1107,
+      "step": 7890
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 8.961559992535922e-06,
+      "loss": 0.1021,
+      "step": 7891
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 8.961426705408793e-06,
+      "loss": 0.1209,
+      "step": 7892
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 8.961293418281664e-06,
+      "loss": 0.1147,
+      "step": 7893
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 8.961160131154535e-06,
+      "loss": 0.1265,
+      "step": 7894
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 8.961026844027406e-06,
+      "loss": 0.1127,
+      "step": 7895
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 8.960893556900275e-06,
+      "loss": 0.1062,
+      "step": 7896
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 8.960760269773146e-06,
+      "loss": 0.0983,
+      "step": 7897
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 8.960626982646017e-06,
+      "loss": 0.0809,
+      "step": 7898
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 8.960493695518888e-06,
+      "loss": 0.1164,
+      "step": 7899
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 8.960360408391759e-06,
+      "loss": 0.1062,
+      "step": 7900
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 8.960227121264628e-06,
+      "loss": 0.1232,
+      "step": 7901
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 8.960093834137499e-06,
+      "loss": 0.1047,
+      "step": 7902
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 8.95996054701037e-06,
+      "loss": 0.097,
+      "step": 7903
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 8.959827259883241e-06,
+      "loss": 0.0996,
+      "step": 7904
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 8.959693972756112e-06,
+      "loss": 0.0989,
+      "step": 7905
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 8.959560685628983e-06,
+      "loss": 0.092,
+      "step": 7906
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 8.959427398501852e-06,
+      "loss": 0.1018,
+      "step": 7907
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 8.959294111374725e-06,
+      "loss": 0.0822,
+      "step": 7908
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 8.959160824247594e-06,
+      "loss": 0.1286,
+      "step": 7909
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 8.959027537120465e-06,
+      "loss": 0.0835,
+      "step": 7910
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 8.958894249993336e-06,
+      "loss": 0.1322,
+      "step": 7911
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 8.958760962866207e-06,
+      "loss": 0.0946,
+      "step": 7912
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 8.958627675739078e-06,
+      "loss": 0.0873,
+      "step": 7913
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 8.95849438861195e-06,
+      "loss": 0.0946,
+      "step": 7914
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 8.958361101484819e-06,
+      "loss": 0.1207,
+      "step": 7915
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 8.958227814357691e-06,
+      "loss": 0.1408,
+      "step": 7916
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 8.95809452723056e-06,
+      "loss": 0.0956,
+      "step": 7917
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 8.957961240103432e-06,
+      "loss": 0.0998,
+      "step": 7918
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 8.957827952976303e-06,
+      "loss": 0.1175,
+      "step": 7919
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 8.957694665849172e-06,
+      "loss": 0.1001,
+      "step": 7920
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 8.957561378722045e-06,
+      "loss": 0.1114,
+      "step": 7921
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 8.957428091594914e-06,
+      "loss": 0.111,
+      "step": 7922
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 8.957294804467785e-06,
+      "loss": 0.1227,
+      "step": 7923
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 8.957161517340656e-06,
+      "loss": 0.151,
+      "step": 7924
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 8.957028230213527e-06,
+      "loss": 0.1458,
+      "step": 7925
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 8.956894943086398e-06,
+      "loss": 0.1188,
+      "step": 7926
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 8.956761655959269e-06,
+      "loss": 0.1237,
+      "step": 7927
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 8.956628368832138e-06,
+      "loss": 0.0854,
+      "step": 7928
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 8.95649508170501e-06,
+      "loss": 0.0853,
+      "step": 7929
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 8.95636179457788e-06,
+      "loss": 0.1113,
+      "step": 7930
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 8.956228507450751e-06,
+      "loss": 0.1216,
+      "step": 7931
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 8.956095220323622e-06,
+      "loss": 0.103,
+      "step": 7932
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 8.955961933196493e-06,
+      "loss": 0.1326,
+      "step": 7933
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 8.955828646069362e-06,
+      "loss": 0.104,
+      "step": 7934
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 8.955695358942235e-06,
+      "loss": 0.0916,
+      "step": 7935
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 8.955562071815104e-06,
+      "loss": 0.1435,
+      "step": 7936
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 8.955428784687975e-06,
+      "loss": 0.1322,
+      "step": 7937
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 8.955295497560846e-06,
+      "loss": 0.0801,
+      "step": 7938
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 8.955162210433716e-06,
+      "loss": 0.1123,
+      "step": 7939
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 8.955028923306588e-06,
+      "loss": 0.0893,
+      "step": 7940
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 8.954895636179458e-06,
+      "loss": 0.1534,
+      "step": 7941
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 8.954762349052329e-06,
+      "loss": 0.0891,
+      "step": 7942
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 8.9546290619252e-06,
+      "loss": 0.1332,
+      "step": 7943
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 8.95449577479807e-06,
+      "loss": 0.1189,
+      "step": 7944
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 8.954362487670942e-06,
+      "loss": 0.1413,
+      "step": 7945
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 8.954229200543813e-06,
+      "loss": 0.0811,
+      "step": 7946
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 8.954095913416682e-06,
+      "loss": 0.0926,
+      "step": 7947
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 8.953962626289555e-06,
+      "loss": 0.0892,
+      "step": 7948
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 8.953829339162424e-06,
+      "loss": 0.1114,
+      "step": 7949
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 8.953696052035295e-06,
+      "loss": 0.1138,
+      "step": 7950
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 8.953562764908166e-06,
+      "loss": 0.1124,
+      "step": 7951
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 8.953429477781037e-06,
+      "loss": 0.0855,
+      "step": 7952
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 8.953296190653908e-06,
+      "loss": 0.0972,
+      "step": 7953
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 8.953162903526779e-06,
+      "loss": 0.1597,
+      "step": 7954
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 8.953029616399648e-06,
+      "loss": 0.1083,
+      "step": 7955
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 8.95289632927252e-06,
+      "loss": 0.1236,
+      "step": 7956
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 8.95276304214539e-06,
+      "loss": 0.0841,
+      "step": 7957
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 8.952629755018261e-06,
+      "loss": 0.1042,
+      "step": 7958
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 8.952496467891132e-06,
+      "loss": 0.1328,
+      "step": 7959
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 8.952363180764002e-06,
+      "loss": 0.0998,
+      "step": 7960
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 8.952229893636874e-06,
+      "loss": 0.1187,
+      "step": 7961
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 8.952096606509744e-06,
+      "loss": 0.0692,
+      "step": 7962
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 8.951963319382615e-06,
+      "loss": 0.1056,
+      "step": 7963
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 8.951830032255486e-06,
+      "loss": 0.122,
+      "step": 7964
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 8.951696745128357e-06,
+      "loss": 0.1337,
+      "step": 7965
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 8.951563458001228e-06,
+      "loss": 0.0993,
+      "step": 7966
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 8.951430170874099e-06,
+      "loss": 0.1186,
+      "step": 7967
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 8.951296883746968e-06,
+      "loss": 0.0653,
+      "step": 7968
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 8.951163596619839e-06,
+      "loss": 0.1172,
+      "step": 7969
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 8.95103030949271e-06,
+      "loss": 0.1386,
+      "step": 7970
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 8.95089702236558e-06,
+      "loss": 0.121,
+      "step": 7971
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 8.950763735238452e-06,
+      "loss": 0.0995,
+      "step": 7972
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 8.950630448111323e-06,
+      "loss": 0.0813,
+      "step": 7973
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 8.950497160984192e-06,
+      "loss": 0.0934,
+      "step": 7974
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 8.950363873857063e-06,
+      "loss": 0.105,
+      "step": 7975
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 8.950230586729934e-06,
+      "loss": 0.1494,
+      "step": 7976
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 8.950097299602805e-06,
+      "loss": 0.1048,
+      "step": 7977
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 8.949964012475676e-06,
+      "loss": 0.1138,
+      "step": 7978
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 8.949830725348545e-06,
+      "loss": 0.1252,
+      "step": 7979
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 8.949697438221418e-06,
+      "loss": 0.143,
+      "step": 7980
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 8.949564151094287e-06,
+      "loss": 0.151,
+      "step": 7981
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 8.949430863967158e-06,
+      "loss": 0.1153,
+      "step": 7982
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 8.94929757684003e-06,
+      "loss": 0.0928,
+      "step": 7983
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 8.9491642897129e-06,
+      "loss": 0.0739,
+      "step": 7984
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 8.949031002585771e-06,
+      "loss": 0.1282,
+      "step": 7985
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 8.948897715458642e-06,
+      "loss": 0.1332,
+      "step": 7986
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 8.948764428331512e-06,
+      "loss": 0.0832,
+      "step": 7987
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 8.948631141204384e-06,
+      "loss": 0.1079,
+      "step": 7988
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 8.948497854077254e-06,
+      "loss": 0.1054,
+      "step": 7989
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 8.948364566950125e-06,
+      "loss": 0.076,
+      "step": 7990
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 8.948231279822996e-06,
+      "loss": 0.0974,
+      "step": 7991
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 8.948097992695867e-06,
+      "loss": 0.1054,
+      "step": 7992
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 8.947964705568738e-06,
+      "loss": 0.1099,
+      "step": 7993
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 8.947831418441609e-06,
+      "loss": 0.0946,
+      "step": 7994
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 8.947698131314478e-06,
+      "loss": 0.1185,
+      "step": 7995
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 8.947564844187349e-06,
+      "loss": 0.1354,
+      "step": 7996
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 8.94743155706022e-06,
+      "loss": 0.1158,
+      "step": 7997
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 8.947298269933091e-06,
+      "loss": 0.1199,
+      "step": 7998
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 8.947164982805962e-06,
+      "loss": 0.0782,
+      "step": 7999
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 8.947031695678831e-06,
+      "loss": 0.1082,
+      "step": 8000
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 8.946898408551702e-06,
+      "loss": 0.0852,
+      "step": 8001
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 8.946765121424573e-06,
+      "loss": 0.1353,
+      "step": 8002
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 8.946631834297444e-06,
+      "loss": 0.0981,
+      "step": 8003
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 8.946498547170315e-06,
+      "loss": 0.1171,
+      "step": 8004
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 8.946365260043186e-06,
+      "loss": 0.081,
+      "step": 8005
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 8.946231972916055e-06,
+      "loss": 0.1393,
+      "step": 8006
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 8.946098685788928e-06,
+      "loss": 0.0803,
+      "step": 8007
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 8.945965398661797e-06,
+      "loss": 0.0987,
+      "step": 8008
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 8.945832111534668e-06,
+      "loss": 0.0988,
+      "step": 8009
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 8.94569882440754e-06,
+      "loss": 0.0913,
+      "step": 8010
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 8.94556553728041e-06,
+      "loss": 0.1359,
+      "step": 8011
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 8.945432250153281e-06,
+      "loss": 0.1117,
+      "step": 8012
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 8.945298963026152e-06,
+      "loss": 0.0624,
+      "step": 8013
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 8.945165675899022e-06,
+      "loss": 0.0847,
+      "step": 8014
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 8.945032388771893e-06,
+      "loss": 0.0966,
+      "step": 8015
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 8.944899101644764e-06,
+      "loss": 0.0908,
+      "step": 8016
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 8.944765814517635e-06,
+      "loss": 0.1417,
+      "step": 8017
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 8.944632527390506e-06,
+      "loss": 0.0865,
+      "step": 8018
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 8.944499240263375e-06,
+      "loss": 0.1013,
+      "step": 8019
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 8.944365953136248e-06,
+      "loss": 0.0964,
+      "step": 8020
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 8.944232666009117e-06,
+      "loss": 0.0868,
+      "step": 8021
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 8.944099378881988e-06,
+      "loss": 0.0663,
+      "step": 8022
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 8.943966091754859e-06,
+      "loss": 0.123,
+      "step": 8023
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 8.94383280462773e-06,
+      "loss": 0.1422,
+      "step": 8024
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 8.943699517500601e-06,
+      "loss": 0.1179,
+      "step": 8025
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 8.943566230373472e-06,
+      "loss": 0.1093,
+      "step": 8026
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 8.943432943246341e-06,
+      "loss": 0.093,
+      "step": 8027
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 8.943299656119214e-06,
+      "loss": 0.1189,
+      "step": 8028
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 8.943166368992083e-06,
+      "loss": 0.092,
+      "step": 8029
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 8.943033081864954e-06,
+      "loss": 0.0983,
+      "step": 8030
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 8.942899794737825e-06,
+      "loss": 0.1009,
+      "step": 8031
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 8.942766507610696e-06,
+      "loss": 0.1255,
+      "step": 8032
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 8.942633220483567e-06,
+      "loss": 0.1428,
+      "step": 8033
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 8.942499933356437e-06,
+      "loss": 0.0938,
+      "step": 8034
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 8.942366646229308e-06,
+      "loss": 0.1391,
+      "step": 8035
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 8.942233359102179e-06,
+      "loss": 0.1138,
+      "step": 8036
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 8.94210007197505e-06,
+      "loss": 0.1142,
+      "step": 8037
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 8.94196678484792e-06,
+      "loss": 0.1081,
+      "step": 8038
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 8.941833497720792e-06,
+      "loss": 0.1047,
+      "step": 8039
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 8.94170021059366e-06,
+      "loss": 0.0986,
+      "step": 8040
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 8.941566923466532e-06,
+      "loss": 0.1025,
+      "step": 8041
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 8.941433636339403e-06,
+      "loss": 0.0845,
+      "step": 8042
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 8.941300349212274e-06,
+      "loss": 0.1361,
+      "step": 8043
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 8.941167062085145e-06,
+      "loss": 0.0786,
+      "step": 8044
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 8.941033774958016e-06,
+      "loss": 0.0771,
+      "step": 8045
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 8.940900487830885e-06,
+      "loss": 0.0881,
+      "step": 8046
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 8.940767200703758e-06,
+      "loss": 0.1185,
+      "step": 8047
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 8.940633913576627e-06,
+      "loss": 0.0967,
+      "step": 8048
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 8.940500626449498e-06,
+      "loss": 0.1049,
+      "step": 8049
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 8.940367339322369e-06,
+      "loss": 0.1148,
+      "step": 8050
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 8.94023405219524e-06,
+      "loss": 0.1024,
+      "step": 8051
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 8.940100765068111e-06,
+      "loss": 0.1049,
+      "step": 8052
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 8.939967477940982e-06,
+      "loss": 0.117,
+      "step": 8053
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 8.939834190813851e-06,
+      "loss": 0.1372,
+      "step": 8054
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 8.939700903686722e-06,
+      "loss": 0.1203,
+      "step": 8055
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 8.939567616559593e-06,
+      "loss": 0.0917,
+      "step": 8056
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 8.939434329432464e-06,
+      "loss": 0.1299,
+      "step": 8057
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 8.939301042305335e-06,
+      "loss": 0.1744,
+      "step": 8058
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 8.939167755178205e-06,
+      "loss": 0.0782,
+      "step": 8059
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 8.939034468051077e-06,
+      "loss": 0.1044,
+      "step": 8060
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 8.938901180923947e-06,
+      "loss": 0.1449,
+      "step": 8061
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 8.938767893796818e-06,
+      "loss": 0.1241,
+      "step": 8062
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 8.938634606669689e-06,
+      "loss": 0.0969,
+      "step": 8063
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 8.93850131954256e-06,
+      "loss": 0.1018,
+      "step": 8064
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 8.93836803241543e-06,
+      "loss": 0.0738,
+      "step": 8065
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 8.938234745288302e-06,
+      "loss": 0.0814,
+      "step": 8066
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 8.938101458161171e-06,
+      "loss": 0.1314,
+      "step": 8067
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 8.937968171034042e-06,
+      "loss": 0.0993,
+      "step": 8068
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 8.937834883906913e-06,
+      "loss": 0.1217,
+      "step": 8069
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 8.937701596779784e-06,
+      "loss": 0.121,
+      "step": 8070
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 8.937568309652655e-06,
+      "loss": 0.0959,
+      "step": 8071
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 8.937435022525526e-06,
+      "loss": 0.1195,
+      "step": 8072
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 8.937301735398395e-06,
+      "loss": 0.0985,
+      "step": 8073
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 8.937168448271266e-06,
+      "loss": 0.0857,
+      "step": 8074
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 8.937035161144137e-06,
+      "loss": 0.085,
+      "step": 8075
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 8.936901874017008e-06,
+      "loss": 0.1099,
+      "step": 8076
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 8.93676858688988e-06,
+      "loss": 0.1059,
+      "step": 8077
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 8.936635299762748e-06,
+      "loss": 0.0812,
+      "step": 8078
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 8.936502012635621e-06,
+      "loss": 0.0809,
+      "step": 8079
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 8.93636872550849e-06,
+      "loss": 0.1011,
+      "step": 8080
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 8.936235438381361e-06,
+      "loss": 0.1183,
+      "step": 8081
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 8.936102151254232e-06,
+      "loss": 0.1042,
+      "step": 8082
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 8.935968864127103e-06,
+      "loss": 0.0865,
+      "step": 8083
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 8.935835576999974e-06,
+      "loss": 0.1328,
+      "step": 8084
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 8.935702289872845e-06,
+      "loss": 0.1402,
+      "step": 8085
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 8.935569002745715e-06,
+      "loss": 0.1144,
+      "step": 8086
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 8.935435715618587e-06,
+      "loss": 0.1294,
+      "step": 8087
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 8.935302428491457e-06,
+      "loss": 0.1034,
+      "step": 8088
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 8.935169141364328e-06,
+      "loss": 0.121,
+      "step": 8089
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 8.935035854237199e-06,
+      "loss": 0.0969,
+      "step": 8090
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 8.93490256711007e-06,
+      "loss": 0.1146,
+      "step": 8091
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 8.93476927998294e-06,
+      "loss": 0.1195,
+      "step": 8092
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 8.93463599285581e-06,
+      "loss": 0.0821,
+      "step": 8093
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 8.934502705728681e-06,
+      "loss": 0.1021,
+      "step": 8094
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 8.934369418601552e-06,
+      "loss": 0.1165,
+      "step": 8095
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 8.934236131474423e-06,
+      "loss": 0.1232,
+      "step": 8096
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 8.934102844347294e-06,
+      "loss": 0.1217,
+      "step": 8097
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 8.933969557220165e-06,
+      "loss": 0.1018,
+      "step": 8098
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 8.933836270093034e-06,
+      "loss": 0.1095,
+      "step": 8099
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 8.933702982965905e-06,
+      "loss": 0.1359,
+      "step": 8100
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 8.933569695838776e-06,
+      "loss": 0.1003,
+      "step": 8101
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 8.933436408711647e-06,
+      "loss": 0.1191,
+      "step": 8102
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 8.933303121584518e-06,
+      "loss": 0.1178,
+      "step": 8103
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 8.93316983445739e-06,
+      "loss": 0.1182,
+      "step": 8104
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 8.933036547330259e-06,
+      "loss": 0.0928,
+      "step": 8105
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 8.932903260203131e-06,
+      "loss": 0.0757,
+      "step": 8106
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 8.932769973076e-06,
+      "loss": 0.0753,
+      "step": 8107
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 8.932636685948872e-06,
+      "loss": 0.1162,
+      "step": 8108
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 8.932503398821743e-06,
+      "loss": 0.1202,
+      "step": 8109
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 8.932370111694614e-06,
+      "loss": 0.0838,
+      "step": 8110
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 8.932236824567485e-06,
+      "loss": 0.1249,
+      "step": 8111
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 8.932103537440354e-06,
+      "loss": 0.201,
+      "step": 8112
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 8.931970250313225e-06,
+      "loss": 0.1182,
+      "step": 8113
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 8.931836963186096e-06,
+      "loss": 0.1387,
+      "step": 8114
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 8.931703676058967e-06,
+      "loss": 0.1092,
+      "step": 8115
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 8.931570388931838e-06,
+      "loss": 0.0832,
+      "step": 8116
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 8.931437101804709e-06,
+      "loss": 0.1069,
+      "step": 8117
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 8.931303814677578e-06,
+      "loss": 0.0849,
+      "step": 8118
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 8.93117052755045e-06,
+      "loss": 0.0801,
+      "step": 8119
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 8.93103724042332e-06,
+      "loss": 0.153,
+      "step": 8120
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 8.930903953296191e-06,
+      "loss": 0.1051,
+      "step": 8121
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 8.930770666169062e-06,
+      "loss": 0.1024,
+      "step": 8122
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 8.930637379041933e-06,
+      "loss": 0.0911,
+      "step": 8123
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 8.930504091914804e-06,
+      "loss": 0.1126,
+      "step": 8124
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 8.930370804787675e-06,
+      "loss": 0.1088,
+      "step": 8125
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 8.930237517660544e-06,
+      "loss": 0.1172,
+      "step": 8126
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 8.930104230533417e-06,
+      "loss": 0.0998,
+      "step": 8127
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 8.929970943406286e-06,
+      "loss": 0.0803,
+      "step": 8128
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 8.929837656279157e-06,
+      "loss": 0.0822,
+      "step": 8129
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 8.929704369152028e-06,
+      "loss": 0.0763,
+      "step": 8130
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 8.9295710820249e-06,
+      "loss": 0.1075,
+      "step": 8131
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 8.92943779489777e-06,
+      "loss": 0.1245,
+      "step": 8132
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 8.92930450777064e-06,
+      "loss": 0.0987,
+      "step": 8133
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 8.92917122064351e-06,
+      "loss": 0.0866,
+      "step": 8134
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 8.929037933516382e-06,
+      "loss": 0.1,
+      "step": 8135
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 8.928904646389253e-06,
+      "loss": 0.0823,
+      "step": 8136
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 8.928771359262124e-06,
+      "loss": 0.0939,
+      "step": 8137
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 8.928638072134995e-06,
+      "loss": 0.1353,
+      "step": 8138
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 8.928504785007864e-06,
+      "loss": 0.0969,
+      "step": 8139
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 8.928371497880735e-06,
+      "loss": 0.1198,
+      "step": 8140
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 8.928238210753606e-06,
+      "loss": 0.0875,
+      "step": 8141
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 8.928104923626477e-06,
+      "loss": 0.087,
+      "step": 8142
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 8.927971636499348e-06,
+      "loss": 0.1449,
+      "step": 8143
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 8.927838349372219e-06,
+      "loss": 0.1395,
+      "step": 8144
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 8.927705062245088e-06,
+      "loss": 0.1143,
+      "step": 8145
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 8.927571775117961e-06,
+      "loss": 0.1225,
+      "step": 8146
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 8.92743848799083e-06,
+      "loss": 0.1168,
+      "step": 8147
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 8.927305200863701e-06,
+      "loss": 0.0896,
+      "step": 8148
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 8.927171913736572e-06,
+      "loss": 0.0749,
+      "step": 8149
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 8.927038626609443e-06,
+      "loss": 0.1322,
+      "step": 8150
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 8.926905339482314e-06,
+      "loss": 0.0767,
+      "step": 8151
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 8.926772052355183e-06,
+      "loss": 0.0888,
+      "step": 8152
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 8.926638765228054e-06,
+      "loss": 0.1312,
+      "step": 8153
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 8.926505478100925e-06,
+      "loss": 0.1234,
+      "step": 8154
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 8.926372190973796e-06,
+      "loss": 0.0798,
+      "step": 8155
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 8.926238903846667e-06,
+      "loss": 0.1201,
+      "step": 8156
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 8.926105616719538e-06,
+      "loss": 0.0873,
+      "step": 8157
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 8.925972329592408e-06,
+      "loss": 0.1258,
+      "step": 8158
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 8.92583904246528e-06,
+      "loss": 0.0733,
+      "step": 8159
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 8.92570575533815e-06,
+      "loss": 0.1541,
+      "step": 8160
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 8.92557246821102e-06,
+      "loss": 0.0774,
+      "step": 8161
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 8.925439181083892e-06,
+      "loss": 0.138,
+      "step": 8162
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 8.925305893956763e-06,
+      "loss": 0.1373,
+      "step": 8163
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 8.925172606829634e-06,
+      "loss": 0.1062,
+      "step": 8164
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 8.925039319702505e-06,
+      "loss": 0.0899,
+      "step": 8165
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 8.924906032575374e-06,
+      "loss": 0.0756,
+      "step": 8166
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 8.924772745448245e-06,
+      "loss": 0.1276,
+      "step": 8167
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 8.924639458321116e-06,
+      "loss": 0.132,
+      "step": 8168
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 8.924506171193987e-06,
+      "loss": 0.0565,
+      "step": 8169
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 8.924372884066858e-06,
+      "loss": 0.111,
+      "step": 8170
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 8.924239596939727e-06,
+      "loss": 0.1345,
+      "step": 8171
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 8.924106309812598e-06,
+      "loss": 0.0889,
+      "step": 8172
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 8.92397302268547e-06,
+      "loss": 0.1006,
+      "step": 8173
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 8.92383973555834e-06,
+      "loss": 0.0674,
+      "step": 8174
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 8.923706448431211e-06,
+      "loss": 0.0975,
+      "step": 8175
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 8.923573161304082e-06,
+      "loss": 0.1103,
+      "step": 8176
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 8.923439874176952e-06,
+      "loss": 0.0861,
+      "step": 8177
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 8.923306587049824e-06,
+      "loss": 0.0869,
+      "step": 8178
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 8.923173299922694e-06,
+      "loss": 0.1379,
+      "step": 8179
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 8.923040012795565e-06,
+      "loss": 0.1002,
+      "step": 8180
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 8.922906725668436e-06,
+      "loss": 0.0978,
+      "step": 8181
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 8.922773438541306e-06,
+      "loss": 0.1052,
+      "step": 8182
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 8.922640151414177e-06,
+      "loss": 0.1262,
+      "step": 8183
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 8.922506864287048e-06,
+      "loss": 0.1183,
+      "step": 8184
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 8.922373577159918e-06,
+      "loss": 0.1119,
+      "step": 8185
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 8.92224029003279e-06,
+      "loss": 0.105,
+      "step": 8186
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 8.92210700290566e-06,
+      "loss": 0.0925,
+      "step": 8187
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 8.92197371577853e-06,
+      "loss": 0.0775,
+      "step": 8188
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 8.921840428651402e-06,
+      "loss": 0.1005,
+      "step": 8189
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 8.921707141524273e-06,
+      "loss": 0.1173,
+      "step": 8190
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 8.921573854397144e-06,
+      "loss": 0.1369,
+      "step": 8191
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 8.921440567270013e-06,
+      "loss": 0.1268,
+      "step": 8192
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 8.921307280142884e-06,
+      "loss": 0.0886,
+      "step": 8193
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 8.921173993015755e-06,
+      "loss": 0.1049,
+      "step": 8194
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 8.921040705888626e-06,
+      "loss": 0.1106,
+      "step": 8195
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 8.920907418761497e-06,
+      "loss": 0.0827,
+      "step": 8196
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 8.920774131634368e-06,
+      "loss": 0.0788,
+      "step": 8197
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 8.920640844507237e-06,
+      "loss": 0.1479,
+      "step": 8198
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 8.92050755738011e-06,
+      "loss": 0.1064,
+      "step": 8199
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 8.92037427025298e-06,
+      "loss": 0.0844,
+      "step": 8200
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 8.92024098312585e-06,
+      "loss": 0.1391,
+      "step": 8201
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 8.920107695998721e-06,
+      "loss": 0.1726,
+      "step": 8202
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 8.919974408871592e-06,
+      "loss": 0.1056,
+      "step": 8203
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 8.919841121744463e-06,
+      "loss": 0.1158,
+      "step": 8204
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 8.919707834617334e-06,
+      "loss": 0.0738,
+      "step": 8205
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 8.919574547490204e-06,
+      "loss": 0.1165,
+      "step": 8206
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 8.919441260363075e-06,
+      "loss": 0.0996,
+      "step": 8207
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 8.919307973235946e-06,
+      "loss": 0.1216,
+      "step": 8208
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 8.919174686108817e-06,
+      "loss": 0.1224,
+      "step": 8209
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 8.919041398981688e-06,
+      "loss": 0.1227,
+      "step": 8210
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 8.918908111854557e-06,
+      "loss": 0.1052,
+      "step": 8211
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 8.918774824727428e-06,
+      "loss": 0.1045,
+      "step": 8212
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 8.918641537600299e-06,
+      "loss": 0.1202,
+      "step": 8213
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 8.91850825047317e-06,
+      "loss": 0.1298,
+      "step": 8214
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 8.918374963346041e-06,
+      "loss": 0.1126,
+      "step": 8215
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 8.918241676218912e-06,
+      "loss": 0.1169,
+      "step": 8216
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 8.918108389091781e-06,
+      "loss": 0.097,
+      "step": 8217
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 8.917975101964654e-06,
+      "loss": 0.1227,
+      "step": 8218
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 8.917841814837523e-06,
+      "loss": 0.1576,
+      "step": 8219
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 8.917708527710394e-06,
+      "loss": 0.1134,
+      "step": 8220
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 8.917575240583265e-06,
+      "loss": 0.0987,
+      "step": 8221
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 8.917441953456136e-06,
+      "loss": 0.097,
+      "step": 8222
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 8.917308666329007e-06,
+      "loss": 0.1194,
+      "step": 8223
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 8.917175379201878e-06,
+      "loss": 0.2009,
+      "step": 8224
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 8.917042092074747e-06,
+      "loss": 0.1053,
+      "step": 8225
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 8.91690880494762e-06,
+      "loss": 0.0944,
+      "step": 8226
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 8.91677551782049e-06,
+      "loss": 0.0989,
+      "step": 8227
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 8.91664223069336e-06,
+      "loss": 0.1137,
+      "step": 8228
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 8.916508943566231e-06,
+      "loss": 0.1266,
+      "step": 8229
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 8.9163756564391e-06,
+      "loss": 0.0961,
+      "step": 8230
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 8.916242369311973e-06,
+      "loss": 0.1157,
+      "step": 8231
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 8.916109082184843e-06,
+      "loss": 0.1209,
+      "step": 8232
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 8.915975795057714e-06,
+      "loss": 0.0971,
+      "step": 8233
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 8.915842507930585e-06,
+      "loss": 0.1133,
+      "step": 8234
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 8.915709220803456e-06,
+      "loss": 0.068,
+      "step": 8235
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 8.915575933676327e-06,
+      "loss": 0.1367,
+      "step": 8236
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 8.915442646549198e-06,
+      "loss": 0.143,
+      "step": 8237
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 8.915309359422067e-06,
+      "loss": 0.1129,
+      "step": 8238
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 8.915176072294938e-06,
+      "loss": 0.1023,
+      "step": 8239
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 8.915042785167809e-06,
+      "loss": 0.1448,
+      "step": 8240
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 8.91490949804068e-06,
+      "loss": 0.1041,
+      "step": 8241
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 8.914776210913551e-06,
+      "loss": 0.1537,
+      "step": 8242
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 8.914642923786422e-06,
+      "loss": 0.09,
+      "step": 8243
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 8.914509636659291e-06,
+      "loss": 0.0934,
+      "step": 8244
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 8.914376349532164e-06,
+      "loss": 0.089,
+      "step": 8245
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 8.914243062405033e-06,
+      "loss": 0.0677,
+      "step": 8246
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 8.914109775277904e-06,
+      "loss": 0.1412,
+      "step": 8247
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 8.913976488150775e-06,
+      "loss": 0.1609,
+      "step": 8248
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 8.913843201023646e-06,
+      "loss": 0.1433,
+      "step": 8249
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 8.913709913896517e-06,
+      "loss": 0.0957,
+      "step": 8250
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 8.913576626769386e-06,
+      "loss": 0.0763,
+      "step": 8251
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 8.913443339642257e-06,
+      "loss": 0.1016,
+      "step": 8252
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 8.913310052515128e-06,
+      "loss": 0.0748,
+      "step": 8253
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 8.913176765388e-06,
+      "loss": 0.0999,
+      "step": 8254
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 8.91304347826087e-06,
+      "loss": 0.0881,
+      "step": 8255
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 8.912910191133741e-06,
+      "loss": 0.0735,
+      "step": 8256
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 8.91277690400661e-06,
+      "loss": 0.0901,
+      "step": 8257
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 8.912643616879483e-06,
+      "loss": 0.0953,
+      "step": 8258
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 8.912510329752353e-06,
+      "loss": 0.1065,
+      "step": 8259
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 8.912377042625224e-06,
+      "loss": 0.117,
+      "step": 8260
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 8.912243755498095e-06,
+      "loss": 0.1025,
+      "step": 8261
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 8.912110468370966e-06,
+      "loss": 0.0822,
+      "step": 8262
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 8.911977181243837e-06,
+      "loss": 0.1645,
+      "step": 8263
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 8.911843894116708e-06,
+      "loss": 0.1204,
+      "step": 8264
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 8.911710606989577e-06,
+      "loss": 0.1091,
+      "step": 8265
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 8.91157731986245e-06,
+      "loss": 0.1199,
+      "step": 8266
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 8.911444032735319e-06,
+      "loss": 0.1148,
+      "step": 8267
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 8.91131074560819e-06,
+      "loss": 0.0943,
+      "step": 8268
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 8.911177458481061e-06,
+      "loss": 0.086,
+      "step": 8269
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 8.91104417135393e-06,
+      "loss": 0.1599,
+      "step": 8270
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 8.910910884226803e-06,
+      "loss": 0.1079,
+      "step": 8271
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 8.910777597099672e-06,
+      "loss": 0.0923,
+      "step": 8272
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 8.910644309972543e-06,
+      "loss": 0.0906,
+      "step": 8273
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 8.910511022845414e-06,
+      "loss": 0.0643,
+      "step": 8274
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 8.910377735718285e-06,
+      "loss": 0.1257,
+      "step": 8275
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 8.910244448591156e-06,
+      "loss": 0.0853,
+      "step": 8276
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 8.910111161464027e-06,
+      "loss": 0.1624,
+      "step": 8277
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 8.909977874336897e-06,
+      "loss": 0.0777,
+      "step": 8278
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 8.909844587209768e-06,
+      "loss": 0.1257,
+      "step": 8279
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 8.909711300082639e-06,
+      "loss": 0.1301,
+      "step": 8280
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 8.90957801295551e-06,
+      "loss": 0.1037,
+      "step": 8281
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 8.90944472582838e-06,
+      "loss": 0.1304,
+      "step": 8282
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 8.909311438701252e-06,
+      "loss": 0.0665,
+      "step": 8283
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 8.909178151574121e-06,
+      "loss": 0.1079,
+      "step": 8284
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 8.909044864446994e-06,
+      "loss": 0.1344,
+      "step": 8285
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 8.908911577319863e-06,
+      "loss": 0.1165,
+      "step": 8286
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 8.908778290192734e-06,
+      "loss": 0.1181,
+      "step": 8287
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 8.908645003065605e-06,
+      "loss": 0.0717,
+      "step": 8288
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 8.908511715938474e-06,
+      "loss": 0.096,
+      "step": 8289
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 8.908378428811347e-06,
+      "loss": 0.1175,
+      "step": 8290
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 8.908245141684216e-06,
+      "loss": 0.1054,
+      "step": 8291
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 8.908111854557087e-06,
+      "loss": 0.0769,
+      "step": 8292
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 8.907978567429958e-06,
+      "loss": 0.109,
+      "step": 8293
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 8.907845280302829e-06,
+      "loss": 0.0772,
+      "step": 8294
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 8.9077119931757e-06,
+      "loss": 0.1097,
+      "step": 8295
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 8.907578706048571e-06,
+      "loss": 0.1126,
+      "step": 8296
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 8.90744541892144e-06,
+      "loss": 0.0922,
+      "step": 8297
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 8.907312131794313e-06,
+      "loss": 0.1079,
+      "step": 8298
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 8.907178844667182e-06,
+      "loss": 0.1245,
+      "step": 8299
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 8.907045557540053e-06,
+      "loss": 0.098,
+      "step": 8300
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 8.906912270412924e-06,
+      "loss": 0.0832,
+      "step": 8301
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 8.906778983285795e-06,
+      "loss": 0.1018,
+      "step": 8302
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 8.906645696158666e-06,
+      "loss": 0.0829,
+      "step": 8303
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 8.906512409031537e-06,
+      "loss": 0.1123,
+      "step": 8304
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 8.906379121904407e-06,
+      "loss": 0.1019,
+      "step": 8305
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 8.906245834777278e-06,
+      "loss": 0.1195,
+      "step": 8306
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 8.906112547650149e-06,
+      "loss": 0.1087,
+      "step": 8307
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 8.90597926052302e-06,
+      "loss": 0.078,
+      "step": 8308
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 8.90584597339589e-06,
+      "loss": 0.1,
+      "step": 8309
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 8.90571268626876e-06,
+      "loss": 0.1024,
+      "step": 8310
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 8.905579399141631e-06,
+      "loss": 0.0931,
+      "step": 8311
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 8.905446112014502e-06,
+      "loss": 0.1193,
+      "step": 8312
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 8.905312824887373e-06,
+      "loss": 0.104,
+      "step": 8313
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 8.905179537760244e-06,
+      "loss": 0.1367,
+      "step": 8314
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 8.905046250633115e-06,
+      "loss": 0.1085,
+      "step": 8315
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 8.904912963505984e-06,
+      "loss": 0.0955,
+      "step": 8316
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 8.904779676378857e-06,
+      "loss": 0.076,
+      "step": 8317
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 8.904646389251726e-06,
+      "loss": 0.1206,
+      "step": 8318
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 8.904513102124597e-06,
+      "loss": 0.0912,
+      "step": 8319
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 8.904379814997468e-06,
+      "loss": 0.1163,
+      "step": 8320
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 8.90424652787034e-06,
+      "loss": 0.0861,
+      "step": 8321
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 8.90411324074321e-06,
+      "loss": 0.081,
+      "step": 8322
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 8.903979953616081e-06,
+      "loss": 0.0834,
+      "step": 8323
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 8.90384666648895e-06,
+      "loss": 0.0935,
+      "step": 8324
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 8.903713379361823e-06,
+      "loss": 0.1196,
+      "step": 8325
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 8.903580092234692e-06,
+      "loss": 0.1224,
+      "step": 8326
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 8.903446805107563e-06,
+      "loss": 0.121,
+      "step": 8327
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 8.903313517980434e-06,
+      "loss": 0.0858,
+      "step": 8328
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 8.903180230853304e-06,
+      "loss": 0.0915,
+      "step": 8329
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 8.903046943726176e-06,
+      "loss": 0.1095,
+      "step": 8330
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 8.902913656599046e-06,
+      "loss": 0.0962,
+      "step": 8331
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 8.902780369471917e-06,
+      "loss": 0.1273,
+      "step": 8332
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 8.902647082344788e-06,
+      "loss": 0.0965,
+      "step": 8333
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 8.902513795217659e-06,
+      "loss": 0.1265,
+      "step": 8334
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 8.90238050809053e-06,
+      "loss": 0.1107,
+      "step": 8335
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 8.9022472209634e-06,
+      "loss": 0.1084,
+      "step": 8336
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 8.90211393383627e-06,
+      "loss": 0.0851,
+      "step": 8337
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 8.901980646709141e-06,
+      "loss": 0.0963,
+      "step": 8338
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 8.901847359582012e-06,
+      "loss": 0.1014,
+      "step": 8339
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 8.901714072454883e-06,
+      "loss": 0.1016,
+      "step": 8340
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 8.901580785327754e-06,
+      "loss": 0.0845,
+      "step": 8341
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 8.901447498200625e-06,
+      "loss": 0.1162,
+      "step": 8342
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 8.901314211073494e-06,
+      "loss": 0.1033,
+      "step": 8343
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 8.901180923946367e-06,
+      "loss": 0.0839,
+      "step": 8344
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 8.901047636819236e-06,
+      "loss": 0.125,
+      "step": 8345
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 8.900914349692107e-06,
+      "loss": 0.138,
+      "step": 8346
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 8.900781062564978e-06,
+      "loss": 0.1017,
+      "step": 8347
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 8.900647775437848e-06,
+      "loss": 0.1323,
+      "step": 8348
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 8.90051448831072e-06,
+      "loss": 0.0913,
+      "step": 8349
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 8.90038120118359e-06,
+      "loss": 0.103,
+      "step": 8350
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 8.90024791405646e-06,
+      "loss": 0.1389,
+      "step": 8351
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 8.900114626929332e-06,
+      "loss": 0.1049,
+      "step": 8352
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 8.899981339802203e-06,
+      "loss": 0.1276,
+      "step": 8353
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 8.899848052675074e-06,
+      "loss": 0.1277,
+      "step": 8354
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 8.899714765547945e-06,
+      "loss": 0.1116,
+      "step": 8355
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 8.899581478420814e-06,
+      "loss": 0.1204,
+      "step": 8356
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 8.899448191293687e-06,
+      "loss": 0.0803,
+      "step": 8357
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 8.899314904166556e-06,
+      "loss": 0.0741,
+      "step": 8358
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 8.899181617039427e-06,
+      "loss": 0.1172,
+      "step": 8359
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 8.899048329912298e-06,
+      "loss": 0.1086,
+      "step": 8360
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 8.898915042785169e-06,
+      "loss": 0.1397,
+      "step": 8361
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 8.89878175565804e-06,
+      "loss": 0.0918,
+      "step": 8362
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 8.89864846853091e-06,
+      "loss": 0.1169,
+      "step": 8363
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 8.89851518140378e-06,
+      "loss": 0.1625,
+      "step": 8364
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 8.898381894276653e-06,
+      "loss": 0.0766,
+      "step": 8365
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 8.898248607149522e-06,
+      "loss": 0.1008,
+      "step": 8366
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 8.898115320022393e-06,
+      "loss": 0.1156,
+      "step": 8367
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 8.897982032895264e-06,
+      "loss": 0.0875,
+      "step": 8368
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 8.897848745768133e-06,
+      "loss": 0.1109,
+      "step": 8369
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 8.897715458641006e-06,
+      "loss": 0.1255,
+      "step": 8370
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 8.897582171513875e-06,
+      "loss": 0.148,
+      "step": 8371
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 8.897448884386746e-06,
+      "loss": 0.0797,
+      "step": 8372
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 8.897315597259617e-06,
+      "loss": 0.0883,
+      "step": 8373
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 8.897182310132488e-06,
+      "loss": 0.136,
+      "step": 8374
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 8.89704902300536e-06,
+      "loss": 0.1251,
+      "step": 8375
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 8.89691573587823e-06,
+      "loss": 0.1095,
+      "step": 8376
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 8.8967824487511e-06,
+      "loss": 0.1288,
+      "step": 8377
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 8.89664916162397e-06,
+      "loss": 0.1185,
+      "step": 8378
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 8.896515874496842e-06,
+      "loss": 0.107,
+      "step": 8379
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 8.896382587369713e-06,
+      "loss": 0.1238,
+      "step": 8380
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 8.896249300242584e-06,
+      "loss": 0.086,
+      "step": 8381
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 8.896116013115455e-06,
+      "loss": 0.1034,
+      "step": 8382
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 8.895982725988324e-06,
+      "loss": 0.1124,
+      "step": 8383
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 8.895849438861197e-06,
+      "loss": 0.1743,
+      "step": 8384
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 8.895716151734066e-06,
+      "loss": 0.0711,
+      "step": 8385
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 8.895582864606937e-06,
+      "loss": 0.0841,
+      "step": 8386
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 8.895449577479808e-06,
+      "loss": 0.1419,
+      "step": 8387
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 8.895316290352677e-06,
+      "loss": 0.1442,
+      "step": 8388
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 8.89518300322555e-06,
+      "loss": 0.07,
+      "step": 8389
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 8.89504971609842e-06,
+      "loss": 0.0801,
+      "step": 8390
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 8.89491642897129e-06,
+      "loss": 0.1045,
+      "step": 8391
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 8.894783141844161e-06,
+      "loss": 0.1055,
+      "step": 8392
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 8.894649854717032e-06,
+      "loss": 0.0985,
+      "step": 8393
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 8.894516567589903e-06,
+      "loss": 0.0723,
+      "step": 8394
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 8.894383280462774e-06,
+      "loss": 0.1821,
+      "step": 8395
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 8.894249993335643e-06,
+      "loss": 0.1078,
+      "step": 8396
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 8.894116706208516e-06,
+      "loss": 0.0846,
+      "step": 8397
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 8.893983419081385e-06,
+      "loss": 0.1146,
+      "step": 8398
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 8.893850131954256e-06,
+      "loss": 0.0981,
+      "step": 8399
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 8.893716844827127e-06,
+      "loss": 0.1273,
+      "step": 8400
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 8.893583557699998e-06,
+      "loss": 0.1148,
+      "step": 8401
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 8.89345027057287e-06,
+      "loss": 0.0928,
+      "step": 8402
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 8.89331698344574e-06,
+      "loss": 0.0889,
+      "step": 8403
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 8.89318369631861e-06,
+      "loss": 0.0946,
+      "step": 8404
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 8.89305040919148e-06,
+      "loss": 0.1478,
+      "step": 8405
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 8.892917122064352e-06,
+      "loss": 0.0923,
+      "step": 8406
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 8.892783834937223e-06,
+      "loss": 0.0894,
+      "step": 8407
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 8.892650547810094e-06,
+      "loss": 0.1421,
+      "step": 8408
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 8.892517260682963e-06,
+      "loss": 0.0689,
+      "step": 8409
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 8.892383973555834e-06,
+      "loss": 0.0766,
+      "step": 8410
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 8.892250686428705e-06,
+      "loss": 0.1016,
+      "step": 8411
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 8.892117399301576e-06,
+      "loss": 0.0962,
+      "step": 8412
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 8.891984112174447e-06,
+      "loss": 0.1039,
+      "step": 8413
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 8.891850825047318e-06,
+      "loss": 0.0991,
+      "step": 8414
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 8.891717537920187e-06,
+      "loss": 0.1347,
+      "step": 8415
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 8.89158425079306e-06,
+      "loss": 0.0894,
+      "step": 8416
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 8.89145096366593e-06,
+      "loss": 0.0719,
+      "step": 8417
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 8.8913176765388e-06,
+      "loss": 0.1267,
+      "step": 8418
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 8.891184389411671e-06,
+      "loss": 0.1123,
+      "step": 8419
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 8.891051102284542e-06,
+      "loss": 0.128,
+      "step": 8420
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 8.890917815157413e-06,
+      "loss": 0.0934,
+      "step": 8421
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 8.890784528030284e-06,
+      "loss": 0.0945,
+      "step": 8422
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 8.890651240903154e-06,
+      "loss": 0.1194,
+      "step": 8423
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 8.890517953776026e-06,
+      "loss": 0.108,
+      "step": 8424
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 8.890384666648896e-06,
+      "loss": 0.1148,
+      "step": 8425
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 8.890251379521767e-06,
+      "loss": 0.1029,
+      "step": 8426
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 8.890118092394638e-06,
+      "loss": 0.1145,
+      "step": 8427
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 8.889984805267507e-06,
+      "loss": 0.1365,
+      "step": 8428
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 8.88985151814038e-06,
+      "loss": 0.0999,
+      "step": 8429
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 8.889718231013249e-06,
+      "loss": 0.1025,
+      "step": 8430
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 8.88958494388612e-06,
+      "loss": 0.089,
+      "step": 8431
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 8.88945165675899e-06,
+      "loss": 0.1281,
+      "step": 8432
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 8.889318369631862e-06,
+      "loss": 0.0978,
+      "step": 8433
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 8.889185082504733e-06,
+      "loss": 0.0858,
+      "step": 8434
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 8.889051795377604e-06,
+      "loss": 0.1144,
+      "step": 8435
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 8.888918508250473e-06,
+      "loss": 0.0954,
+      "step": 8436
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 8.888785221123346e-06,
+      "loss": 0.1483,
+      "step": 8437
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 8.888651933996215e-06,
+      "loss": 0.1688,
+      "step": 8438
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 8.888518646869086e-06,
+      "loss": 0.0741,
+      "step": 8439
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 8.888385359741957e-06,
+      "loss": 0.1443,
+      "step": 8440
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 8.888252072614828e-06,
+      "loss": 0.123,
+      "step": 8441
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 8.888118785487699e-06,
+      "loss": 0.1335,
+      "step": 8442
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 8.88798549836057e-06,
+      "loss": 0.0999,
+      "step": 8443
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 8.88785221123344e-06,
+      "loss": 0.088,
+      "step": 8444
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 8.88771892410631e-06,
+      "loss": 0.119,
+      "step": 8445
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 8.887585636979181e-06,
+      "loss": 0.1004,
+      "step": 8446
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 8.887452349852052e-06,
+      "loss": 0.1112,
+      "step": 8447
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 8.887319062724923e-06,
+      "loss": 0.0914,
+      "step": 8448
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 8.887185775597793e-06,
+      "loss": 0.111,
+      "step": 8449
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 8.887052488470664e-06,
+      "loss": 0.1089,
+      "step": 8450
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 8.886919201343535e-06,
+      "loss": 0.1062,
+      "step": 8451
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 8.886785914216406e-06,
+      "loss": 0.0918,
+      "step": 8452
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 8.886652627089277e-06,
+      "loss": 0.1002,
+      "step": 8453
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 8.886519339962148e-06,
+      "loss": 0.0679,
+      "step": 8454
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 8.886386052835017e-06,
+      "loss": 0.1208,
+      "step": 8455
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 8.88625276570789e-06,
+      "loss": 0.118,
+      "step": 8456
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 8.886119478580759e-06,
+      "loss": 0.118,
+      "step": 8457
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 8.88598619145363e-06,
+      "loss": 0.1105,
+      "step": 8458
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 8.885852904326501e-06,
+      "loss": 0.114,
+      "step": 8459
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 8.885719617199372e-06,
+      "loss": 0.1285,
+      "step": 8460
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 8.885586330072243e-06,
+      "loss": 0.1411,
+      "step": 8461
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 8.885453042945114e-06,
+      "loss": 0.1265,
+      "step": 8462
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 8.885319755817983e-06,
+      "loss": 0.116,
+      "step": 8463
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 8.885186468690856e-06,
+      "loss": 0.0878,
+      "step": 8464
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 8.885053181563725e-06,
+      "loss": 0.094,
+      "step": 8465
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 8.884919894436596e-06,
+      "loss": 0.094,
+      "step": 8466
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 8.884786607309467e-06,
+      "loss": 0.1047,
+      "step": 8467
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 8.884653320182336e-06,
+      "loss": 0.1583,
+      "step": 8468
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 8.884520033055209e-06,
+      "loss": 0.0685,
+      "step": 8469
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 8.884386745928078e-06,
+      "loss": 0.067,
+      "step": 8470
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 8.88425345880095e-06,
+      "loss": 0.1057,
+      "step": 8471
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 8.88412017167382e-06,
+      "loss": 0.1079,
+      "step": 8472
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 8.883986884546691e-06,
+      "loss": 0.09,
+      "step": 8473
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 8.883853597419562e-06,
+      "loss": 0.0694,
+      "step": 8474
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 8.883720310292433e-06,
+      "loss": 0.1002,
+      "step": 8475
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 8.883587023165303e-06,
+      "loss": 0.0904,
+      "step": 8476
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 8.883453736038174e-06,
+      "loss": 0.1086,
+      "step": 8477
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 8.883320448911045e-06,
+      "loss": 0.0935,
+      "step": 8478
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 8.883187161783916e-06,
+      "loss": 0.0873,
+      "step": 8479
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 8.883053874656787e-06,
+      "loss": 0.0893,
+      "step": 8480
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 8.882920587529658e-06,
+      "loss": 0.0934,
+      "step": 8481
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 8.882787300402527e-06,
+      "loss": 0.1087,
+      "step": 8482
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 8.8826540132754e-06,
+      "loss": 0.1014,
+      "step": 8483
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 8.882520726148269e-06,
+      "loss": 0.1157,
+      "step": 8484
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 8.88238743902114e-06,
+      "loss": 0.1047,
+      "step": 8485
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 8.882254151894011e-06,
+      "loss": 0.1118,
+      "step": 8486
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 8.88212086476688e-06,
+      "loss": 0.151,
+      "step": 8487
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 8.881987577639753e-06,
+      "loss": 0.1008,
+      "step": 8488
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 8.881854290512622e-06,
+      "loss": 0.1261,
+      "step": 8489
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 8.881721003385493e-06,
+      "loss": 0.1153,
+      "step": 8490
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 8.881587716258364e-06,
+      "loss": 0.1375,
+      "step": 8491
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 8.881454429131235e-06,
+      "loss": 0.101,
+      "step": 8492
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 8.881321142004106e-06,
+      "loss": 0.092,
+      "step": 8493
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 8.881187854876977e-06,
+      "loss": 0.0837,
+      "step": 8494
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 8.881054567749847e-06,
+      "loss": 0.0946,
+      "step": 8495
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 8.88092128062272e-06,
+      "loss": 0.0781,
+      "step": 8496
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 8.880787993495589e-06,
+      "loss": 0.1385,
+      "step": 8497
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 8.88065470636846e-06,
+      "loss": 0.1083,
+      "step": 8498
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 8.88052141924133e-06,
+      "loss": 0.1082,
+      "step": 8499
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 8.880388132114202e-06,
+      "loss": 0.0913,
+      "step": 8500
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 8.880254844987072e-06,
+      "loss": 0.1152,
+      "step": 8501
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 8.880121557859943e-06,
+      "loss": 0.1078,
+      "step": 8502
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 8.879988270732813e-06,
+      "loss": 0.0875,
+      "step": 8503
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 8.879854983605684e-06,
+      "loss": 0.1032,
+      "step": 8504
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 8.879721696478555e-06,
+      "loss": 0.0701,
+      "step": 8505
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 8.879588409351426e-06,
+      "loss": 0.0637,
+      "step": 8506
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 8.879455122224297e-06,
+      "loss": 0.1095,
+      "step": 8507
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 8.879321835097166e-06,
+      "loss": 0.1251,
+      "step": 8508
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 8.879188547970039e-06,
+      "loss": 0.0894,
+      "step": 8509
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 8.879055260842908e-06,
+      "loss": 0.0963,
+      "step": 8510
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 8.878921973715779e-06,
+      "loss": 0.1102,
+      "step": 8511
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 8.87878868658865e-06,
+      "loss": 0.0849,
+      "step": 8512
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 8.878655399461521e-06,
+      "loss": 0.0729,
+      "step": 8513
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 8.878522112334392e-06,
+      "loss": 0.1213,
+      "step": 8514
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 8.878388825207263e-06,
+      "loss": 0.0852,
+      "step": 8515
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 8.878255538080132e-06,
+      "loss": 0.0931,
+      "step": 8516
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 8.878122250953003e-06,
+      "loss": 0.099,
+      "step": 8517
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 8.877988963825874e-06,
+      "loss": 0.1107,
+      "step": 8518
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 8.877855676698745e-06,
+      "loss": 0.0974,
+      "step": 8519
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 8.877722389571616e-06,
+      "loss": 0.0627,
+      "step": 8520
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 8.877589102444487e-06,
+      "loss": 0.1174,
+      "step": 8521
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 8.877455815317357e-06,
+      "loss": 0.0893,
+      "step": 8522
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 8.87732252819023e-06,
+      "loss": 0.1238,
+      "step": 8523
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 8.877189241063099e-06,
+      "loss": 0.0893,
+      "step": 8524
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 8.87705595393597e-06,
+      "loss": 0.1246,
+      "step": 8525
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 8.87692266680884e-06,
+      "loss": 0.107,
+      "step": 8526
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 8.87678937968171e-06,
+      "loss": 0.0749,
+      "step": 8527
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 8.876656092554583e-06,
+      "loss": 0.1226,
+      "step": 8528
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 8.876522805427452e-06,
+      "loss": 0.0742,
+      "step": 8529
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 8.876389518300323e-06,
+      "loss": 0.0858,
+      "step": 8530
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 8.876256231173194e-06,
+      "loss": 0.1128,
+      "step": 8531
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 8.876122944046065e-06,
+      "loss": 0.0775,
+      "step": 8532
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 8.875989656918936e-06,
+      "loss": 0.0855,
+      "step": 8533
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 8.875856369791807e-06,
+      "loss": 0.1295,
+      "step": 8534
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 8.875723082664676e-06,
+      "loss": 0.095,
+      "step": 8535
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 8.875589795537549e-06,
+      "loss": 0.1012,
+      "step": 8536
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 8.875456508410418e-06,
+      "loss": 0.1203,
+      "step": 8537
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 8.875323221283289e-06,
+      "loss": 0.0935,
+      "step": 8538
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 8.87518993415616e-06,
+      "loss": 0.0899,
+      "step": 8539
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 8.875056647029031e-06,
+      "loss": 0.1438,
+      "step": 8540
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 8.874923359901902e-06,
+      "loss": 0.0916,
+      "step": 8541
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 8.874790072774773e-06,
+      "loss": 0.1244,
+      "step": 8542
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 8.874656785647642e-06,
+      "loss": 0.1396,
+      "step": 8543
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 8.874523498520513e-06,
+      "loss": 0.1186,
+      "step": 8544
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 8.874390211393384e-06,
+      "loss": 0.0738,
+      "step": 8545
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 8.874256924266255e-06,
+      "loss": 0.1449,
+      "step": 8546
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 8.874123637139126e-06,
+      "loss": 0.1099,
+      "step": 8547
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 8.873990350011996e-06,
+      "loss": 0.0796,
+      "step": 8548
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 8.873857062884867e-06,
+      "loss": 0.0952,
+      "step": 8549
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 8.873723775757738e-06,
+      "loss": 0.122,
+      "step": 8550
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 8.873590488630609e-06,
+      "loss": 0.0888,
+      "step": 8551
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 8.87345720150348e-06,
+      "loss": 0.0953,
+      "step": 8552
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 8.87332391437635e-06,
+      "loss": 0.0743,
+      "step": 8553
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 8.87319062724922e-06,
+      "loss": 0.112,
+      "step": 8554
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 8.873057340122093e-06,
+      "loss": 0.0816,
+      "step": 8555
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 8.872924052994962e-06,
+      "loss": 0.0657,
+      "step": 8556
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 8.872790765867833e-06,
+      "loss": 0.0812,
+      "step": 8557
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 8.872657478740704e-06,
+      "loss": 0.0936,
+      "step": 8558
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 8.872524191613575e-06,
+      "loss": 0.1139,
+      "step": 8559
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 8.872390904486446e-06,
+      "loss": 0.1377,
+      "step": 8560
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 8.872257617359317e-06,
+      "loss": 0.1266,
+      "step": 8561
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 8.872124330232186e-06,
+      "loss": 0.0844,
+      "step": 8562
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 8.871991043105057e-06,
+      "loss": 0.1373,
+      "step": 8563
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 8.871857755977928e-06,
+      "loss": 0.1551,
+      "step": 8564
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 8.8717244688508e-06,
+      "loss": 0.1275,
+      "step": 8565
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 8.87159118172367e-06,
+      "loss": 0.1061,
+      "step": 8566
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 8.87145789459654e-06,
+      "loss": 0.1432,
+      "step": 8567
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 8.871324607469412e-06,
+      "loss": 0.1224,
+      "step": 8568
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 8.871191320342282e-06,
+      "loss": 0.0605,
+      "step": 8569
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 8.871058033215152e-06,
+      "loss": 0.1422,
+      "step": 8570
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 8.870924746088023e-06,
+      "loss": 0.0995,
+      "step": 8571
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 8.870791458960894e-06,
+      "loss": 0.0767,
+      "step": 8572
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 8.870658171833765e-06,
+      "loss": 0.1074,
+      "step": 8573
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 8.870524884706636e-06,
+      "loss": 0.1304,
+      "step": 8574
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 8.870391597579506e-06,
+      "loss": 0.188,
+      "step": 8575
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 8.870258310452378e-06,
+      "loss": 0.1154,
+      "step": 8576
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 8.870125023325248e-06,
+      "loss": 0.0807,
+      "step": 8577
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 8.869991736198119e-06,
+      "loss": 0.1499,
+      "step": 8578
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 8.86985844907099e-06,
+      "loss": 0.1159,
+      "step": 8579
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 8.86972516194386e-06,
+      "loss": 0.0788,
+      "step": 8580
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 8.86959187481673e-06,
+      "loss": 0.0736,
+      "step": 8581
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 8.869458587689603e-06,
+      "loss": 0.1207,
+      "step": 8582
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 8.869325300562472e-06,
+      "loss": 0.105,
+      "step": 8583
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 8.869192013435343e-06,
+      "loss": 0.1893,
+      "step": 8584
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 8.869058726308214e-06,
+      "loss": 0.1399,
+      "step": 8585
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 8.868925439181083e-06,
+      "loss": 0.1191,
+      "step": 8586
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 8.868792152053956e-06,
+      "loss": 0.125,
+      "step": 8587
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 8.868658864926825e-06,
+      "loss": 0.0941,
+      "step": 8588
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 8.868525577799696e-06,
+      "loss": 0.1217,
+      "step": 8589
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 8.868392290672567e-06,
+      "loss": 0.108,
+      "step": 8590
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 8.868259003545438e-06,
+      "loss": 0.1175,
+      "step": 8591
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 8.86812571641831e-06,
+      "loss": 0.1158,
+      "step": 8592
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 8.86799242929118e-06,
+      "loss": 0.1012,
+      "step": 8593
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 8.86785914216405e-06,
+      "loss": 0.0824,
+      "step": 8594
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 8.867725855036922e-06,
+      "loss": 0.122,
+      "step": 8595
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 8.867592567909792e-06,
+      "loss": 0.1537,
+      "step": 8596
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 8.867459280782663e-06,
+      "loss": 0.1316,
+      "step": 8597
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 8.867325993655534e-06,
+      "loss": 0.0833,
+      "step": 8598
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 8.867192706528405e-06,
+      "loss": 0.105,
+      "step": 8599
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 8.867059419401276e-06,
+      "loss": 0.1469,
+      "step": 8600
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 8.866926132274147e-06,
+      "loss": 0.1032,
+      "step": 8601
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 8.866792845147016e-06,
+      "loss": 0.0979,
+      "step": 8602
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 8.866659558019887e-06,
+      "loss": 0.0939,
+      "step": 8603
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 8.866526270892758e-06,
+      "loss": 0.1014,
+      "step": 8604
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 8.866392983765629e-06,
+      "loss": 0.1085,
+      "step": 8605
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 8.8662596966385e-06,
+      "loss": 0.1075,
+      "step": 8606
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 8.866126409511369e-06,
+      "loss": 0.1077,
+      "step": 8607
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 8.865993122384242e-06,
+      "loss": 0.1047,
+      "step": 8608
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 8.865859835257111e-06,
+      "loss": 0.1367,
+      "step": 8609
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 8.865726548129982e-06,
+      "loss": 0.0964,
+      "step": 8610
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 8.865593261002853e-06,
+      "loss": 0.106,
+      "step": 8611
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 8.865459973875724e-06,
+      "loss": 0.1059,
+      "step": 8612
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 8.865326686748595e-06,
+      "loss": 0.1309,
+      "step": 8613
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 8.865193399621466e-06,
+      "loss": 0.0913,
+      "step": 8614
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 8.865060112494335e-06,
+      "loss": 0.1238,
+      "step": 8615
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 8.864926825367206e-06,
+      "loss": 0.104,
+      "step": 8616
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 8.864793538240077e-06,
+      "loss": 0.0738,
+      "step": 8617
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 8.864660251112948e-06,
+      "loss": 0.0999,
+      "step": 8618
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 8.86452696398582e-06,
+      "loss": 0.1181,
+      "step": 8619
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 8.86439367685869e-06,
+      "loss": 0.1038,
+      "step": 8620
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 8.86426038973156e-06,
+      "loss": 0.1108,
+      "step": 8621
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 8.86412710260443e-06,
+      "loss": 0.1357,
+      "step": 8622
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 8.863993815477302e-06,
+      "loss": 0.0836,
+      "step": 8623
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 8.863860528350173e-06,
+      "loss": 0.1261,
+      "step": 8624
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 8.863727241223044e-06,
+      "loss": 0.1048,
+      "step": 8625
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 8.863593954095913e-06,
+      "loss": 0.0633,
+      "step": 8626
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 8.863460666968786e-06,
+      "loss": 0.0811,
+      "step": 8627
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 8.863327379841655e-06,
+      "loss": 0.0961,
+      "step": 8628
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 8.863194092714526e-06,
+      "loss": 0.147,
+      "step": 8629
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 8.863060805587397e-06,
+      "loss": 0.1238,
+      "step": 8630
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 8.862927518460268e-06,
+      "loss": 0.1031,
+      "step": 8631
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 8.862794231333139e-06,
+      "loss": 0.1137,
+      "step": 8632
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 8.86266094420601e-06,
+      "loss": 0.1503,
+      "step": 8633
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 8.86252765707888e-06,
+      "loss": 0.101,
+      "step": 8634
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 8.862394369951752e-06,
+      "loss": 0.1119,
+      "step": 8635
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 8.862261082824621e-06,
+      "loss": 0.1138,
+      "step": 8636
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 8.862127795697492e-06,
+      "loss": 0.1071,
+      "step": 8637
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 8.861994508570363e-06,
+      "loss": 0.1401,
+      "step": 8638
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 8.861861221443234e-06,
+      "loss": 0.1077,
+      "step": 8639
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 8.861727934316105e-06,
+      "loss": 0.1286,
+      "step": 8640
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 8.861594647188974e-06,
+      "loss": 0.0862,
+      "step": 8641
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 8.861461360061845e-06,
+      "loss": 0.1262,
+      "step": 8642
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 8.861328072934716e-06,
+      "loss": 0.1014,
+      "step": 8643
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 8.861194785807587e-06,
+      "loss": 0.0971,
+      "step": 8644
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 8.861061498680458e-06,
+      "loss": 0.0929,
+      "step": 8645
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 8.86092821155333e-06,
+      "loss": 0.1169,
+      "step": 8646
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 8.860794924426199e-06,
+      "loss": 0.0941,
+      "step": 8647
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 8.86066163729907e-06,
+      "loss": 0.1669,
+      "step": 8648
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 8.86052835017194e-06,
+      "loss": 0.0783,
+      "step": 8649
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 8.860395063044812e-06,
+      "loss": 0.1034,
+      "step": 8650
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 8.860261775917683e-06,
+      "loss": 0.0887,
+      "step": 8651
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 8.860128488790554e-06,
+      "loss": 0.0578,
+      "step": 8652
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 8.859995201663423e-06,
+      "loss": 0.1169,
+      "step": 8653
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 8.859861914536296e-06,
+      "loss": 0.1099,
+      "step": 8654
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 8.859728627409165e-06,
+      "loss": 0.105,
+      "step": 8655
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 8.859595340282036e-06,
+      "loss": 0.0957,
+      "step": 8656
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 8.859462053154907e-06,
+      "loss": 0.1332,
+      "step": 8657
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 8.859328766027778e-06,
+      "loss": 0.0814,
+      "step": 8658
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 8.859195478900649e-06,
+      "loss": 0.0916,
+      "step": 8659
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 8.85906219177352e-06,
+      "loss": 0.1086,
+      "step": 8660
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 8.85892890464639e-06,
+      "loss": 0.0725,
+      "step": 8661
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 8.85879561751926e-06,
+      "loss": 0.1099,
+      "step": 8662
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 8.858662330392131e-06,
+      "loss": 0.1182,
+      "step": 8663
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 8.858529043265002e-06,
+      "loss": 0.1185,
+      "step": 8664
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 8.858395756137873e-06,
+      "loss": 0.1245,
+      "step": 8665
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 8.858262469010743e-06,
+      "loss": 0.0602,
+      "step": 8666
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 8.858129181883615e-06,
+      "loss": 0.1194,
+      "step": 8667
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 8.857995894756485e-06,
+      "loss": 0.109,
+      "step": 8668
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 8.857862607629356e-06,
+      "loss": 0.1429,
+      "step": 8669
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 8.857729320502227e-06,
+      "loss": 0.1075,
+      "step": 8670
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 8.857596033375098e-06,
+      "loss": 0.1068,
+      "step": 8671
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 8.857462746247969e-06,
+      "loss": 0.0949,
+      "step": 8672
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 8.85732945912084e-06,
+      "loss": 0.1161,
+      "step": 8673
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 8.857196171993709e-06,
+      "loss": 0.1013,
+      "step": 8674
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 8.857062884866582e-06,
+      "loss": 0.0912,
+      "step": 8675
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 8.85692959773945e-06,
+      "loss": 0.1072,
+      "step": 8676
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 8.856796310612322e-06,
+      "loss": 0.0832,
+      "step": 8677
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 8.856663023485193e-06,
+      "loss": 0.1317,
+      "step": 8678
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 8.856529736358064e-06,
+      "loss": 0.0756,
+      "step": 8679
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 8.856396449230935e-06,
+      "loss": 0.0961,
+      "step": 8680
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 8.856263162103804e-06,
+      "loss": 0.0819,
+      "step": 8681
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 8.856129874976675e-06,
+      "loss": 0.0872,
+      "step": 8682
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 8.855996587849546e-06,
+      "loss": 0.0758,
+      "step": 8683
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 8.855863300722417e-06,
+      "loss": 0.1119,
+      "step": 8684
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 8.855730013595288e-06,
+      "loss": 0.1402,
+      "step": 8685
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 8.855596726468159e-06,
+      "loss": 0.0898,
+      "step": 8686
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 8.855463439341028e-06,
+      "loss": 0.117,
+      "step": 8687
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 8.8553301522139e-06,
+      "loss": 0.164,
+      "step": 8688
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 8.85519686508677e-06,
+      "loss": 0.0975,
+      "step": 8689
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 8.855063577959641e-06,
+      "loss": 0.0912,
+      "step": 8690
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 8.854930290832512e-06,
+      "loss": 0.1105,
+      "step": 8691
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 8.854797003705383e-06,
+      "loss": 0.1033,
+      "step": 8692
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 8.854663716578253e-06,
+      "loss": 0.1169,
+      "step": 8693
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 8.854530429451125e-06,
+      "loss": 0.1156,
+      "step": 8694
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 8.854397142323995e-06,
+      "loss": 0.1384,
+      "step": 8695
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 8.854263855196866e-06,
+      "loss": 0.1758,
+      "step": 8696
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 8.854130568069737e-06,
+      "loss": 0.1293,
+      "step": 8697
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 8.853997280942608e-06,
+      "loss": 0.0948,
+      "step": 8698
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 8.853863993815479e-06,
+      "loss": 0.0959,
+      "step": 8699
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 8.853730706688348e-06,
+      "loss": 0.094,
+      "step": 8700
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 8.853597419561219e-06,
+      "loss": 0.0871,
+      "step": 8701
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 8.85346413243409e-06,
+      "loss": 0.0786,
+      "step": 8702
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 8.853330845306961e-06,
+      "loss": 0.1042,
+      "step": 8703
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 8.853197558179832e-06,
+      "loss": 0.0953,
+      "step": 8704
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 8.853064271052703e-06,
+      "loss": 0.1112,
+      "step": 8705
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 8.852930983925572e-06,
+      "loss": 0.0843,
+      "step": 8706
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 8.852797696798445e-06,
+      "loss": 0.0995,
+      "step": 8707
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 8.852664409671314e-06,
+      "loss": 0.1108,
+      "step": 8708
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 8.852531122544185e-06,
+      "loss": 0.1111,
+      "step": 8709
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 8.852397835417056e-06,
+      "loss": 0.0939,
+      "step": 8710
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 8.852264548289927e-06,
+      "loss": 0.0842,
+      "step": 8711
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 8.852131261162798e-06,
+      "loss": 0.1012,
+      "step": 8712
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 8.85199797403567e-06,
+      "loss": 0.1136,
+      "step": 8713
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 8.851864686908538e-06,
+      "loss": 0.118,
+      "step": 8714
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 8.85173139978141e-06,
+      "loss": 0.1308,
+      "step": 8715
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 8.85159811265428e-06,
+      "loss": 0.0916,
+      "step": 8716
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 8.851464825527151e-06,
+      "loss": 0.1337,
+      "step": 8717
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 8.851331538400022e-06,
+      "loss": 0.1097,
+      "step": 8718
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 8.851198251272893e-06,
+      "loss": 0.1273,
+      "step": 8719
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 8.851064964145763e-06,
+      "loss": 0.0931,
+      "step": 8720
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 8.850931677018634e-06,
+      "loss": 0.1486,
+      "step": 8721
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 8.850798389891505e-06,
+      "loss": 0.1074,
+      "step": 8722
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 8.850665102764376e-06,
+      "loss": 0.0745,
+      "step": 8723
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 8.850531815637247e-06,
+      "loss": 0.115,
+      "step": 8724
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 8.850398528510116e-06,
+      "loss": 0.1109,
+      "step": 8725
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 8.850265241382989e-06,
+      "loss": 0.105,
+      "step": 8726
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 8.850131954255858e-06,
+      "loss": 0.1088,
+      "step": 8727
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 8.849998667128729e-06,
+      "loss": 0.0985,
+      "step": 8728
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 8.8498653800016e-06,
+      "loss": 0.0953,
+      "step": 8729
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 8.849732092874471e-06,
+      "loss": 0.1357,
+      "step": 8730
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 8.849598805747342e-06,
+      "loss": 0.1429,
+      "step": 8731
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 8.849465518620213e-06,
+      "loss": 0.062,
+      "step": 8732
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 8.849332231493082e-06,
+      "loss": 0.1081,
+      "step": 8733
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 8.849198944365955e-06,
+      "loss": 0.1052,
+      "step": 8734
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 8.849065657238824e-06,
+      "loss": 0.0974,
+      "step": 8735
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 8.848932370111695e-06,
+      "loss": 0.1185,
+      "step": 8736
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 8.848799082984566e-06,
+      "loss": 0.1062,
+      "step": 8737
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 8.848665795857437e-06,
+      "loss": 0.1278,
+      "step": 8738
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 8.848532508730308e-06,
+      "loss": 0.0976,
+      "step": 8739
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 8.848399221603178e-06,
+      "loss": 0.1239,
+      "step": 8740
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 8.848265934476049e-06,
+      "loss": 0.077,
+      "step": 8741
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 8.84813264734892e-06,
+      "loss": 0.1085,
+      "step": 8742
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 8.84799936022179e-06,
+      "loss": 0.1493,
+      "step": 8743
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 8.847866073094662e-06,
+      "loss": 0.0962,
+      "step": 8744
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 8.847732785967533e-06,
+      "loss": 0.0917,
+      "step": 8745
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 8.847599498840402e-06,
+      "loss": 0.0976,
+      "step": 8746
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 8.847466211713275e-06,
+      "loss": 0.1515,
+      "step": 8747
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 8.847332924586144e-06,
+      "loss": 0.1256,
+      "step": 8748
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 8.847199637459015e-06,
+      "loss": 0.1352,
+      "step": 8749
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 8.847066350331886e-06,
+      "loss": 0.0835,
+      "step": 8750
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 8.846933063204757e-06,
+      "loss": 0.0668,
+      "step": 8751
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 8.846799776077628e-06,
+      "loss": 0.0865,
+      "step": 8752
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 8.846666488950499e-06,
+      "loss": 0.1231,
+      "step": 8753
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 8.846533201823368e-06,
+      "loss": 0.1848,
+      "step": 8754
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 8.846399914696239e-06,
+      "loss": 0.093,
+      "step": 8755
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 8.84626662756911e-06,
+      "loss": 0.113,
+      "step": 8756
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 8.846133340441981e-06,
+      "loss": 0.1452,
+      "step": 8757
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 8.846000053314852e-06,
+      "loss": 0.1222,
+      "step": 8758
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 8.845866766187721e-06,
+      "loss": 0.0783,
+      "step": 8759
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 8.845733479060592e-06,
+      "loss": 0.1351,
+      "step": 8760
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 8.845600191933463e-06,
+      "loss": 0.0927,
+      "step": 8761
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 8.845466904806334e-06,
+      "loss": 0.1085,
+      "step": 8762
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 8.845333617679205e-06,
+      "loss": 0.1192,
+      "step": 8763
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 8.845200330552076e-06,
+      "loss": 0.1191,
+      "step": 8764
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 8.845067043424946e-06,
+      "loss": 0.0983,
+      "step": 8765
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 8.844933756297818e-06,
+      "loss": 0.0811,
+      "step": 8766
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 8.844800469170688e-06,
+      "loss": 0.0735,
+      "step": 8767
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 8.844667182043559e-06,
+      "loss": 0.1126,
+      "step": 8768
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 8.84453389491643e-06,
+      "loss": 0.1356,
+      "step": 8769
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 8.8444006077893e-06,
+      "loss": 0.1013,
+      "step": 8770
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 8.844267320662172e-06,
+      "loss": 0.0996,
+      "step": 8771
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 8.844134033535043e-06,
+      "loss": 0.0851,
+      "step": 8772
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 8.844000746407912e-06,
+      "loss": 0.0776,
+      "step": 8773
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 8.843867459280785e-06,
+      "loss": 0.1208,
+      "step": 8774
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 8.843734172153654e-06,
+      "loss": 0.0851,
+      "step": 8775
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 8.843600885026525e-06,
+      "loss": 0.1249,
+      "step": 8776
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 8.843467597899396e-06,
+      "loss": 0.138,
+      "step": 8777
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 8.843334310772265e-06,
+      "loss": 0.096,
+      "step": 8778
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 8.843201023645138e-06,
+      "loss": 0.0939,
+      "step": 8779
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 8.843067736518007e-06,
+      "loss": 0.0923,
+      "step": 8780
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 8.842934449390878e-06,
+      "loss": 0.1412,
+      "step": 8781
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 8.84280116226375e-06,
+      "loss": 0.1322,
+      "step": 8782
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 8.84266787513662e-06,
+      "loss": 0.0865,
+      "step": 8783
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 8.842534588009491e-06,
+      "loss": 0.0935,
+      "step": 8784
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 8.842401300882362e-06,
+      "loss": 0.0972,
+      "step": 8785
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 8.842268013755231e-06,
+      "loss": 0.0728,
+      "step": 8786
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 8.842134726628102e-06,
+      "loss": 0.0976,
+      "step": 8787
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 8.842001439500973e-06,
+      "loss": 0.0967,
+      "step": 8788
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 8.841868152373844e-06,
+      "loss": 0.1121,
+      "step": 8789
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 8.841734865246715e-06,
+      "loss": 0.0713,
+      "step": 8790
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 8.841601578119586e-06,
+      "loss": 0.0987,
+      "step": 8791
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 8.841468290992456e-06,
+      "loss": 0.1061,
+      "step": 8792
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 8.841335003865328e-06,
+      "loss": 0.1106,
+      "step": 8793
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 8.841201716738198e-06,
+      "loss": 0.1095,
+      "step": 8794
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 8.841068429611069e-06,
+      "loss": 0.112,
+      "step": 8795
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 8.84093514248394e-06,
+      "loss": 0.085,
+      "step": 8796
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 8.84080185535681e-06,
+      "loss": 0.1121,
+      "step": 8797
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 8.840668568229682e-06,
+      "loss": 0.0943,
+      "step": 8798
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 8.840535281102551e-06,
+      "loss": 0.1064,
+      "step": 8799
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 8.840401993975422e-06,
+      "loss": 0.1106,
+      "step": 8800
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 8.840268706848293e-06,
+      "loss": 0.1211,
+      "step": 8801
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 8.840135419721164e-06,
+      "loss": 0.178,
+      "step": 8802
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 8.840002132594035e-06,
+      "loss": 0.1393,
+      "step": 8803
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 8.839868845466906e-06,
+      "loss": 0.1018,
+      "step": 8804
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 8.839735558339775e-06,
+      "loss": 0.1091,
+      "step": 8805
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 8.839602271212648e-06,
+      "loss": 0.0966,
+      "step": 8806
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 8.839468984085517e-06,
+      "loss": 0.139,
+      "step": 8807
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 8.839335696958388e-06,
+      "loss": 0.1212,
+      "step": 8808
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 8.83920240983126e-06,
+      "loss": 0.1023,
+      "step": 8809
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 8.83906912270413e-06,
+      "loss": 0.0786,
+      "step": 8810
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 8.838935835577001e-06,
+      "loss": 0.0859,
+      "step": 8811
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 8.838802548449872e-06,
+      "loss": 0.12,
+      "step": 8812
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 8.838669261322742e-06,
+      "loss": 0.0989,
+      "step": 8813
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 8.838535974195614e-06,
+      "loss": 0.1202,
+      "step": 8814
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 8.838402687068484e-06,
+      "loss": 0.0609,
+      "step": 8815
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 8.838269399941355e-06,
+      "loss": 0.1401,
+      "step": 8816
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 8.838136112814226e-06,
+      "loss": 0.1108,
+      "step": 8817
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 8.838002825687095e-06,
+      "loss": 0.0993,
+      "step": 8818
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 8.837869538559967e-06,
+      "loss": 0.0878,
+      "step": 8819
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 8.837736251432837e-06,
+      "loss": 0.0722,
+      "step": 8820
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 8.837602964305708e-06,
+      "loss": 0.1421,
+      "step": 8821
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 8.837469677178579e-06,
+      "loss": 0.0614,
+      "step": 8822
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 8.83733639005145e-06,
+      "loss": 0.1361,
+      "step": 8823
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 8.83720310292432e-06,
+      "loss": 0.0839,
+      "step": 8824
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 8.837069815797192e-06,
+      "loss": 0.0728,
+      "step": 8825
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 8.836936528670061e-06,
+      "loss": 0.1109,
+      "step": 8826
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 8.836803241542932e-06,
+      "loss": 0.114,
+      "step": 8827
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 8.836669954415803e-06,
+      "loss": 0.1225,
+      "step": 8828
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 8.836536667288674e-06,
+      "loss": 0.147,
+      "step": 8829
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 8.836403380161545e-06,
+      "loss": 0.1355,
+      "step": 8830
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 8.836270093034416e-06,
+      "loss": 0.1009,
+      "step": 8831
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 8.836136805907285e-06,
+      "loss": 0.1157,
+      "step": 8832
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 8.836003518780158e-06,
+      "loss": 0.0861,
+      "step": 8833
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 8.835870231653027e-06,
+      "loss": 0.1067,
+      "step": 8834
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 8.835736944525898e-06,
+      "loss": 0.0672,
+      "step": 8835
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 8.83560365739877e-06,
+      "loss": 0.0755,
+      "step": 8836
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 8.835470370271639e-06,
+      "loss": 0.1026,
+      "step": 8837
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 8.835337083144511e-06,
+      "loss": 0.1196,
+      "step": 8838
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 8.83520379601738e-06,
+      "loss": 0.0713,
+      "step": 8839
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 8.835070508890252e-06,
+      "loss": 0.1182,
+      "step": 8840
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 8.834937221763123e-06,
+      "loss": 0.0913,
+      "step": 8841
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 8.834803934635994e-06,
+      "loss": 0.091,
+      "step": 8842
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 8.834670647508865e-06,
+      "loss": 0.0876,
+      "step": 8843
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 8.834537360381736e-06,
+      "loss": 0.108,
+      "step": 8844
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 8.834404073254605e-06,
+      "loss": 0.1174,
+      "step": 8845
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 8.834270786127478e-06,
+      "loss": 0.1151,
+      "step": 8846
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 8.834137499000347e-06,
+      "loss": 0.1002,
+      "step": 8847
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 8.834004211873218e-06,
+      "loss": 0.1338,
+      "step": 8848
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 8.833870924746089e-06,
+      "loss": 0.1032,
+      "step": 8849
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 8.83373763761896e-06,
+      "loss": 0.1304,
+      "step": 8850
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 8.833604350491831e-06,
+      "loss": 0.0878,
+      "step": 8851
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 8.833471063364702e-06,
+      "loss": 0.0963,
+      "step": 8852
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 8.833337776237571e-06,
+      "loss": 0.107,
+      "step": 8853
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 8.833204489110442e-06,
+      "loss": 0.103,
+      "step": 8854
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 8.833071201983313e-06,
+      "loss": 0.1086,
+      "step": 8855
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 8.832937914856184e-06,
+      "loss": 0.1261,
+      "step": 8856
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 8.832804627729055e-06,
+      "loss": 0.0886,
+      "step": 8857
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 8.832671340601924e-06,
+      "loss": 0.0895,
+      "step": 8858
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 8.832538053474795e-06,
+      "loss": 0.1135,
+      "step": 8859
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 8.832404766347666e-06,
+      "loss": 0.1383,
+      "step": 8860
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 8.832271479220537e-06,
+      "loss": 0.1065,
+      "step": 8861
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 8.832138192093408e-06,
+      "loss": 0.0731,
+      "step": 8862
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 8.83200490496628e-06,
+      "loss": 0.1249,
+      "step": 8863
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 8.831871617839149e-06,
+      "loss": 0.1003,
+      "step": 8864
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 8.831738330712021e-06,
+      "loss": 0.1035,
+      "step": 8865
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 8.83160504358489e-06,
+      "loss": 0.0616,
+      "step": 8866
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 8.831471756457762e-06,
+      "loss": 0.0903,
+      "step": 8867
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 8.831338469330633e-06,
+      "loss": 0.0996,
+      "step": 8868
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 8.831205182203504e-06,
+      "loss": 0.1299,
+      "step": 8869
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 8.831071895076375e-06,
+      "loss": 0.0853,
+      "step": 8870
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 8.830938607949246e-06,
+      "loss": 0.1004,
+      "step": 8871
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 8.830805320822115e-06,
+      "loss": 0.0882,
+      "step": 8872
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 8.830672033694988e-06,
+      "loss": 0.0978,
+      "step": 8873
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 8.830538746567857e-06,
+      "loss": 0.1348,
+      "step": 8874
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 8.830405459440728e-06,
+      "loss": 0.0924,
+      "step": 8875
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 8.830272172313599e-06,
+      "loss": 0.1098,
+      "step": 8876
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 8.830138885186468e-06,
+      "loss": 0.1004,
+      "step": 8877
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 8.830005598059341e-06,
+      "loss": 0.1305,
+      "step": 8878
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 8.82987231093221e-06,
+      "loss": 0.1587,
+      "step": 8879
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 8.829739023805081e-06,
+      "loss": 0.1307,
+      "step": 8880
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 8.829605736677952e-06,
+      "loss": 0.1229,
+      "step": 8881
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 8.829472449550823e-06,
+      "loss": 0.1152,
+      "step": 8882
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 8.829339162423694e-06,
+      "loss": 0.1641,
+      "step": 8883
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 8.829205875296565e-06,
+      "loss": 0.1133,
+      "step": 8884
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 8.829072588169435e-06,
+      "loss": 0.1974,
+      "step": 8885
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 8.828939301042306e-06,
+      "loss": 0.1343,
+      "step": 8886
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 8.828806013915177e-06,
+      "loss": 0.1083,
+      "step": 8887
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 8.828672726788047e-06,
+      "loss": 0.0762,
+      "step": 8888
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 8.828539439660918e-06,
+      "loss": 0.1265,
+      "step": 8889
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 8.82840615253379e-06,
+      "loss": 0.1296,
+      "step": 8890
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 8.828272865406659e-06,
+      "loss": 0.0908,
+      "step": 8891
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 8.828139578279531e-06,
+      "loss": 0.094,
+      "step": 8892
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 8.8280062911524e-06,
+      "loss": 0.1257,
+      "step": 8893
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 8.827873004025272e-06,
+      "loss": 0.1074,
+      "step": 8894
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 8.827739716898143e-06,
+      "loss": 0.0812,
+      "step": 8895
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 8.827606429771012e-06,
+      "loss": 0.1329,
+      "step": 8896
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 8.827473142643885e-06,
+      "loss": 0.0952,
+      "step": 8897
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 8.827339855516754e-06,
+      "loss": 0.1053,
+      "step": 8898
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 8.827206568389625e-06,
+      "loss": 0.0912,
+      "step": 8899
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 8.827073281262496e-06,
+      "loss": 0.1045,
+      "step": 8900
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 8.826939994135367e-06,
+      "loss": 0.0666,
+      "step": 8901
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 8.826806707008238e-06,
+      "loss": 0.0714,
+      "step": 8902
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 8.826673419881109e-06,
+      "loss": 0.0884,
+      "step": 8903
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 8.826540132753978e-06,
+      "loss": 0.0803,
+      "step": 8904
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 8.826406845626851e-06,
+      "loss": 0.088,
+      "step": 8905
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 8.82627355849972e-06,
+      "loss": 0.0902,
+      "step": 8906
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 8.826140271372591e-06,
+      "loss": 0.0907,
+      "step": 8907
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 8.826006984245462e-06,
+      "loss": 0.1056,
+      "step": 8908
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 8.825873697118333e-06,
+      "loss": 0.0783,
+      "step": 8909
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 8.825740409991204e-06,
+      "loss": 0.0987,
+      "step": 8910
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 8.825607122864075e-06,
+      "loss": 0.0779,
+      "step": 8911
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 8.825473835736945e-06,
+      "loss": 0.0765,
+      "step": 8912
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 8.825340548609817e-06,
+      "loss": 0.0742,
+      "step": 8913
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 8.825207261482687e-06,
+      "loss": 0.1256,
+      "step": 8914
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 8.825073974355558e-06,
+      "loss": 0.1526,
+      "step": 8915
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 8.824940687228429e-06,
+      "loss": 0.1078,
+      "step": 8916
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 8.824807400101298e-06,
+      "loss": 0.1012,
+      "step": 8917
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 8.82467411297417e-06,
+      "loss": 0.0779,
+      "step": 8918
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 8.82454082584704e-06,
+      "loss": 0.1005,
+      "step": 8919
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 8.824407538719911e-06,
+      "loss": 0.0693,
+      "step": 8920
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 8.824274251592782e-06,
+      "loss": 0.0993,
+      "step": 8921
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 8.824140964465653e-06,
+      "loss": 0.1042,
+      "step": 8922
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 8.824007677338524e-06,
+      "loss": 0.0804,
+      "step": 8923
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 8.823874390211395e-06,
+      "loss": 0.1098,
+      "step": 8924
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 8.823741103084264e-06,
+      "loss": 0.1245,
+      "step": 8925
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 8.823607815957135e-06,
+      "loss": 0.0811,
+      "step": 8926
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 8.823474528830006e-06,
+      "loss": 0.0652,
+      "step": 8927
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 8.823341241702877e-06,
+      "loss": 0.1079,
+      "step": 8928
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 8.823207954575748e-06,
+      "loss": 0.0756,
+      "step": 8929
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 8.823074667448619e-06,
+      "loss": 0.1275,
+      "step": 8930
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 8.822941380321488e-06,
+      "loss": 0.0966,
+      "step": 8931
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 8.822808093194361e-06,
+      "loss": 0.0725,
+      "step": 8932
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 8.82267480606723e-06,
+      "loss": 0.0925,
+      "step": 8933
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 8.822541518940101e-06,
+      "loss": 0.1104,
+      "step": 8934
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 8.822408231812972e-06,
+      "loss": 0.1194,
+      "step": 8935
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 8.822274944685842e-06,
+      "loss": 0.1056,
+      "step": 8936
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 8.822141657558714e-06,
+      "loss": 0.1214,
+      "step": 8937
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 8.822008370431584e-06,
+      "loss": 0.0916,
+      "step": 8938
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 8.821875083304455e-06,
+      "loss": 0.0915,
+      "step": 8939
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 8.821741796177326e-06,
+      "loss": 0.1028,
+      "step": 8940
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 8.821608509050197e-06,
+      "loss": 0.1398,
+      "step": 8941
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 8.821475221923068e-06,
+      "loss": 0.0732,
+      "step": 8942
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 8.821341934795939e-06,
+      "loss": 0.1019,
+      "step": 8943
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 8.821208647668808e-06,
+      "loss": 0.163,
+      "step": 8944
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 8.82107536054168e-06,
+      "loss": 0.0772,
+      "step": 8945
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 8.82094207341455e-06,
+      "loss": 0.0973,
+      "step": 8946
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 8.820808786287421e-06,
+      "loss": 0.1018,
+      "step": 8947
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 8.820675499160292e-06,
+      "loss": 0.1086,
+      "step": 8948
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 8.820542212033163e-06,
+      "loss": 0.1225,
+      "step": 8949
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 8.820408924906034e-06,
+      "loss": 0.1215,
+      "step": 8950
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 8.820275637778905e-06,
+      "loss": 0.1004,
+      "step": 8951
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 8.820142350651774e-06,
+      "loss": 0.1269,
+      "step": 8952
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 8.820009063524645e-06,
+      "loss": 0.0869,
+      "step": 8953
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 8.819875776397516e-06,
+      "loss": 0.1203,
+      "step": 8954
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 8.819742489270387e-06,
+      "loss": 0.1465,
+      "step": 8955
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 8.819609202143258e-06,
+      "loss": 0.1328,
+      "step": 8956
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 8.819475915016127e-06,
+      "loss": 0.0938,
+      "step": 8957
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 8.819342627888998e-06,
+      "loss": 0.1156,
+      "step": 8958
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 8.81920934076187e-06,
+      "loss": 0.1363,
+      "step": 8959
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 8.81907605363474e-06,
+      "loss": 0.0933,
+      "step": 8960
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 8.818942766507611e-06,
+      "loss": 0.1017,
+      "step": 8961
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 8.818809479380482e-06,
+      "loss": 0.0999,
+      "step": 8962
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 8.818676192253352e-06,
+      "loss": 0.0893,
+      "step": 8963
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 8.818542905126224e-06,
+      "loss": 0.0795,
+      "step": 8964
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 8.818409617999094e-06,
+      "loss": 0.107,
+      "step": 8965
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 8.818276330871965e-06,
+      "loss": 0.0882,
+      "step": 8966
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 8.818143043744836e-06,
+      "loss": 0.1184,
+      "step": 8967
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 8.818009756617707e-06,
+      "loss": 0.1608,
+      "step": 8968
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 8.817876469490578e-06,
+      "loss": 0.1819,
+      "step": 8969
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 8.817743182363449e-06,
+      "loss": 0.1064,
+      "step": 8970
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 8.817609895236318e-06,
+      "loss": 0.0807,
+      "step": 8971
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 8.81747660810919e-06,
+      "loss": 0.0852,
+      "step": 8972
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 8.81734332098206e-06,
+      "loss": 0.0906,
+      "step": 8973
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 8.817210033854931e-06,
+      "loss": 0.1105,
+      "step": 8974
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 8.817076746727802e-06,
+      "loss": 0.1041,
+      "step": 8975
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 8.816943459600671e-06,
+      "loss": 0.1298,
+      "step": 8976
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 8.816810172473544e-06,
+      "loss": 0.1089,
+      "step": 8977
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 8.816676885346413e-06,
+      "loss": 0.0767,
+      "step": 8978
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 8.816543598219284e-06,
+      "loss": 0.0863,
+      "step": 8979
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 8.816410311092155e-06,
+      "loss": 0.0736,
+      "step": 8980
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 8.816277023965026e-06,
+      "loss": 0.1382,
+      "step": 8981
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 8.816143736837897e-06,
+      "loss": 0.1089,
+      "step": 8982
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 8.816010449710768e-06,
+      "loss": 0.1259,
+      "step": 8983
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 8.815877162583638e-06,
+      "loss": 0.1569,
+      "step": 8984
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 8.81574387545651e-06,
+      "loss": 0.1001,
+      "step": 8985
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 8.81561058832938e-06,
+      "loss": 0.1068,
+      "step": 8986
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 8.81547730120225e-06,
+      "loss": 0.0772,
+      "step": 8987
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 8.815344014075122e-06,
+      "loss": 0.0761,
+      "step": 8988
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 8.815210726947993e-06,
+      "loss": 0.1378,
+      "step": 8989
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 8.815077439820864e-06,
+      "loss": 0.0667,
+      "step": 8990
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 8.814944152693735e-06,
+      "loss": 0.1369,
+      "step": 8991
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 8.814810865566604e-06,
+      "loss": 0.1507,
+      "step": 8992
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 8.814677578439475e-06,
+      "loss": 0.0726,
+      "step": 8993
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 8.814544291312346e-06,
+      "loss": 0.0902,
+      "step": 8994
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 8.814411004185217e-06,
+      "loss": 0.131,
+      "step": 8995
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 8.814277717058088e-06,
+      "loss": 0.1089,
+      "step": 8996
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 8.814144429930957e-06,
+      "loss": 0.0897,
+      "step": 8997
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 8.814011142803828e-06,
+      "loss": 0.0902,
+      "step": 8998
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 8.813877855676699e-06,
+      "loss": 0.1258,
+      "step": 8999
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 8.81374456854957e-06,
+      "loss": 0.0801,
+      "step": 9000
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 8.813611281422441e-06,
+      "loss": 0.09,
+      "step": 9001
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 8.813477994295312e-06,
+      "loss": 0.1103,
+      "step": 9002
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 8.813344707168181e-06,
+      "loss": 0.0805,
+      "step": 9003
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 8.813211420041054e-06,
+      "loss": 0.0868,
+      "step": 9004
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 8.813078132913923e-06,
+      "loss": 0.1559,
+      "step": 9005
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 8.812944845786794e-06,
+      "loss": 0.1049,
+      "step": 9006
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 8.812811558659665e-06,
+      "loss": 0.1076,
+      "step": 9007
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 8.812678271532536e-06,
+      "loss": 0.0979,
+      "step": 9008
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 8.812544984405407e-06,
+      "loss": 0.1083,
+      "step": 9009
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 8.812411697278278e-06,
+      "loss": 0.1088,
+      "step": 9010
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 8.812278410151148e-06,
+      "loss": 0.0883,
+      "step": 9011
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 8.81214512302402e-06,
+      "loss": 0.112,
+      "step": 9012
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 8.81201183589689e-06,
+      "loss": 0.1283,
+      "step": 9013
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 8.81187854876976e-06,
+      "loss": 0.086,
+      "step": 9014
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 8.811745261642632e-06,
+      "loss": 0.0847,
+      "step": 9015
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 8.811611974515501e-06,
+      "loss": 0.1059,
+      "step": 9016
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 8.811478687388374e-06,
+      "loss": 0.0887,
+      "step": 9017
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 8.811345400261243e-06,
+      "loss": 0.107,
+      "step": 9018
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 8.811212113134114e-06,
+      "loss": 0.1047,
+      "step": 9019
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 8.811078826006985e-06,
+      "loss": 0.1197,
+      "step": 9020
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 8.810945538879856e-06,
+      "loss": 0.0912,
+      "step": 9021
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 8.810812251752727e-06,
+      "loss": 0.1205,
+      "step": 9022
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 8.810678964625598e-06,
+      "loss": 0.0884,
+      "step": 9023
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 8.810545677498467e-06,
+      "loss": 0.1135,
+      "step": 9024
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 8.810412390371338e-06,
+      "loss": 0.1116,
+      "step": 9025
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 8.81027910324421e-06,
+      "loss": 0.1024,
+      "step": 9026
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 8.81014581611708e-06,
+      "loss": 0.1206,
+      "step": 9027
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 8.810012528989951e-06,
+      "loss": 0.1239,
+      "step": 9028
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 8.809879241862822e-06,
+      "loss": 0.1204,
+      "step": 9029
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 8.809745954735691e-06,
+      "loss": 0.0987,
+      "step": 9030
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 8.809612667608564e-06,
+      "loss": 0.1181,
+      "step": 9031
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 8.809479380481433e-06,
+      "loss": 0.1321,
+      "step": 9032
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 8.809346093354304e-06,
+      "loss": 0.1106,
+      "step": 9033
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 8.809212806227175e-06,
+      "loss": 0.1723,
+      "step": 9034
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 8.809079519100045e-06,
+      "loss": 0.1066,
+      "step": 9035
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 8.808946231972917e-06,
+      "loss": 0.1428,
+      "step": 9036
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 8.808812944845787e-06,
+      "loss": 0.1065,
+      "step": 9037
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 8.808679657718658e-06,
+      "loss": 0.1323,
+      "step": 9038
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 8.808546370591529e-06,
+      "loss": 0.0917,
+      "step": 9039
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 8.8084130834644e-06,
+      "loss": 0.1136,
+      "step": 9040
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 8.80827979633727e-06,
+      "loss": 0.1366,
+      "step": 9041
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 8.808146509210142e-06,
+      "loss": 0.0651,
+      "step": 9042
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 8.808013222083011e-06,
+      "loss": 0.0788,
+      "step": 9043
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 8.807879934955884e-06,
+      "loss": 0.0903,
+      "step": 9044
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 8.807746647828753e-06,
+      "loss": 0.0892,
+      "step": 9045
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 8.807613360701624e-06,
+      "loss": 0.0964,
+      "step": 9046
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 8.807480073574495e-06,
+      "loss": 0.0938,
+      "step": 9047
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 8.807346786447366e-06,
+      "loss": 0.1484,
+      "step": 9048
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 8.807213499320237e-06,
+      "loss": 0.0932,
+      "step": 9049
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 8.807080212193108e-06,
+      "loss": 0.0941,
+      "step": 9050
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 8.806946925065977e-06,
+      "loss": 0.1085,
+      "step": 9051
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 8.80681363793885e-06,
+      "loss": 0.1186,
+      "step": 9052
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 8.80668035081172e-06,
+      "loss": 0.0908,
+      "step": 9053
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 8.80654706368459e-06,
+      "loss": 0.1244,
+      "step": 9054
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 8.806413776557461e-06,
+      "loss": 0.0891,
+      "step": 9055
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 8.80628048943033e-06,
+      "loss": 0.1368,
+      "step": 9056
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 8.806147202303203e-06,
+      "loss": 0.1405,
+      "step": 9057
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 8.806013915176073e-06,
+      "loss": 0.1141,
+      "step": 9058
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 8.805880628048944e-06,
+      "loss": 0.0982,
+      "step": 9059
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 8.805747340921815e-06,
+      "loss": 0.0685,
+      "step": 9060
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 8.805614053794686e-06,
+      "loss": 0.0846,
+      "step": 9061
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 8.805480766667557e-06,
+      "loss": 0.0773,
+      "step": 9062
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 8.805347479540428e-06,
+      "loss": 0.0938,
+      "step": 9063
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 8.805214192413297e-06,
+      "loss": 0.1079,
+      "step": 9064
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 8.805080905286168e-06,
+      "loss": 0.0915,
+      "step": 9065
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 8.804947618159039e-06,
+      "loss": 0.0575,
+      "step": 9066
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 8.80481433103191e-06,
+      "loss": 0.1107,
+      "step": 9067
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 8.80468104390478e-06,
+      "loss": 0.0906,
+      "step": 9068
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 8.804547756777652e-06,
+      "loss": 0.0745,
+      "step": 9069
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 8.804414469650521e-06,
+      "loss": 0.0807,
+      "step": 9070
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 8.804281182523394e-06,
+      "loss": 0.1455,
+      "step": 9071
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 8.804147895396263e-06,
+      "loss": 0.0998,
+      "step": 9072
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 8.804014608269134e-06,
+      "loss": 0.0776,
+      "step": 9073
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 8.803881321142005e-06,
+      "loss": 0.0847,
+      "step": 9074
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 8.803748034014874e-06,
+      "loss": 0.1028,
+      "step": 9075
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 8.803614746887747e-06,
+      "loss": 0.0972,
+      "step": 9076
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 8.803481459760616e-06,
+      "loss": 0.0947,
+      "step": 9077
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 8.803348172633487e-06,
+      "loss": 0.1117,
+      "step": 9078
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 8.803214885506358e-06,
+      "loss": 0.1105,
+      "step": 9079
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 8.80308159837923e-06,
+      "loss": 0.1257,
+      "step": 9080
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 8.8029483112521e-06,
+      "loss": 0.1089,
+      "step": 9081
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 8.802815024124971e-06,
+      "loss": 0.118,
+      "step": 9082
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 8.80268173699784e-06,
+      "loss": 0.1276,
+      "step": 9083
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 8.802548449870713e-06,
+      "loss": 0.0772,
+      "step": 9084
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 8.802415162743583e-06,
+      "loss": 0.1334,
+      "step": 9085
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 8.802281875616454e-06,
+      "loss": 0.1116,
+      "step": 9086
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 8.802148588489325e-06,
+      "loss": 0.0989,
+      "step": 9087
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 8.802015301362196e-06,
+      "loss": 0.0736,
+      "step": 9088
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 8.801882014235067e-06,
+      "loss": 0.0919,
+      "step": 9089
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 8.801748727107938e-06,
+      "loss": 0.1153,
+      "step": 9090
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 8.801615439980807e-06,
+      "loss": 0.0935,
+      "step": 9091
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 8.801482152853678e-06,
+      "loss": 0.1137,
+      "step": 9092
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 8.801348865726549e-06,
+      "loss": 0.0993,
+      "step": 9093
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 8.80121557859942e-06,
+      "loss": 0.1562,
+      "step": 9094
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 8.801082291472291e-06,
+      "loss": 0.0798,
+      "step": 9095
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 8.80094900434516e-06,
+      "loss": 0.0745,
+      "step": 9096
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 8.800815717218031e-06,
+      "loss": 0.0986,
+      "step": 9097
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 8.800682430090902e-06,
+      "loss": 0.0707,
+      "step": 9098
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 8.800549142963773e-06,
+      "loss": 0.0798,
+      "step": 9099
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 8.800415855836644e-06,
+      "loss": 0.1432,
+      "step": 9100
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 8.800282568709515e-06,
+      "loss": 0.1274,
+      "step": 9101
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 8.800149281582384e-06,
+      "loss": 0.1062,
+      "step": 9102
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 8.800015994455257e-06,
+      "loss": 0.0651,
+      "step": 9103
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 8.799882707328126e-06,
+      "loss": 0.1023,
+      "step": 9104
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 8.799749420200997e-06,
+      "loss": 0.1348,
+      "step": 9105
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 8.799616133073868e-06,
+      "loss": 0.0853,
+      "step": 9106
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 8.79948284594674e-06,
+      "loss": 0.1346,
+      "step": 9107
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 8.79934955881961e-06,
+      "loss": 0.0967,
+      "step": 9108
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 8.799216271692481e-06,
+      "loss": 0.1369,
+      "step": 9109
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 8.79908298456535e-06,
+      "loss": 0.1324,
+      "step": 9110
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 8.798949697438222e-06,
+      "loss": 0.1398,
+      "step": 9111
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 8.798816410311093e-06,
+      "loss": 0.1072,
+      "step": 9112
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 8.798683123183964e-06,
+      "loss": 0.0892,
+      "step": 9113
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 8.798549836056835e-06,
+      "loss": 0.1231,
+      "step": 9114
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 8.798416548929704e-06,
+      "loss": 0.1238,
+      "step": 9115
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 8.798283261802577e-06,
+      "loss": 0.0786,
+      "step": 9116
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 8.798149974675446e-06,
+      "loss": 0.052,
+      "step": 9117
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 8.798016687548317e-06,
+      "loss": 0.0948,
+      "step": 9118
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 8.797883400421188e-06,
+      "loss": 0.0955,
+      "step": 9119
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 8.797750113294059e-06,
+      "loss": 0.0966,
+      "step": 9120
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 8.79761682616693e-06,
+      "loss": 0.0713,
+      "step": 9121
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 8.797483539039801e-06,
+      "loss": 0.1656,
+      "step": 9122
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 8.79735025191267e-06,
+      "loss": 0.1,
+      "step": 9123
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 8.797216964785541e-06,
+      "loss": 0.1523,
+      "step": 9124
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 8.797083677658412e-06,
+      "loss": 0.1097,
+      "step": 9125
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 8.796950390531283e-06,
+      "loss": 0.0965,
+      "step": 9126
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 8.796817103404154e-06,
+      "loss": 0.0948,
+      "step": 9127
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 8.796683816277025e-06,
+      "loss": 0.082,
+      "step": 9128
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 8.796550529149895e-06,
+      "loss": 0.0941,
+      "step": 9129
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 8.796417242022767e-06,
+      "loss": 0.0889,
+      "step": 9130
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 8.796283954895637e-06,
+      "loss": 0.1026,
+      "step": 9131
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 8.796150667768508e-06,
+      "loss": 0.1681,
+      "step": 9132
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 8.796017380641379e-06,
+      "loss": 0.1027,
+      "step": 9133
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 8.795884093514248e-06,
+      "loss": 0.1104,
+      "step": 9134
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 8.79575080638712e-06,
+      "loss": 0.0697,
+      "step": 9135
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 8.79561751925999e-06,
+      "loss": 0.1162,
+      "step": 9136
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 8.79548423213286e-06,
+      "loss": 0.116,
+      "step": 9137
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 8.795350945005732e-06,
+      "loss": 0.1432,
+      "step": 9138
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 8.795217657878603e-06,
+      "loss": 0.1231,
+      "step": 9139
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 8.795084370751474e-06,
+      "loss": 0.098,
+      "step": 9140
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 8.794951083624345e-06,
+      "loss": 0.1082,
+      "step": 9141
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 8.794817796497214e-06,
+      "loss": 0.1144,
+      "step": 9142
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 8.794684509370087e-06,
+      "loss": 0.0992,
+      "step": 9143
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 8.794551222242956e-06,
+      "loss": 0.117,
+      "step": 9144
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 8.794417935115827e-06,
+      "loss": 0.1022,
+      "step": 9145
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 8.794284647988698e-06,
+      "loss": 0.0614,
+      "step": 9146
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 8.794151360861569e-06,
+      "loss": 0.133,
+      "step": 9147
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 8.79401807373444e-06,
+      "loss": 0.1587,
+      "step": 9148
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 8.793884786607311e-06,
+      "loss": 0.1328,
+      "step": 9149
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 8.79375149948018e-06,
+      "loss": 0.1026,
+      "step": 9150
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 8.793618212353051e-06,
+      "loss": 0.0748,
+      "step": 9151
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 8.793484925225922e-06,
+      "loss": 0.1281,
+      "step": 9152
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 8.793351638098793e-06,
+      "loss": 0.0918,
+      "step": 9153
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 8.793218350971664e-06,
+      "loss": 0.1196,
+      "step": 9154
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 8.793085063844534e-06,
+      "loss": 0.1415,
+      "step": 9155
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 8.792951776717406e-06,
+      "loss": 0.0959,
+      "step": 9156
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 8.792818489590276e-06,
+      "loss": 0.0974,
+      "step": 9157
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 8.792685202463147e-06,
+      "loss": 0.1245,
+      "step": 9158
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 8.792551915336018e-06,
+      "loss": 0.0969,
+      "step": 9159
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 8.792418628208889e-06,
+      "loss": 0.1303,
+      "step": 9160
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 8.79228534108176e-06,
+      "loss": 0.0783,
+      "step": 9161
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 8.79215205395463e-06,
+      "loss": 0.077,
+      "step": 9162
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 8.7920187668275e-06,
+      "loss": 0.0775,
+      "step": 9163
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 8.791885479700371e-06,
+      "loss": 0.09,
+      "step": 9164
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 8.791752192573242e-06,
+      "loss": 0.0862,
+      "step": 9165
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 8.791618905446113e-06,
+      "loss": 0.1618,
+      "step": 9166
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 8.791485618318984e-06,
+      "loss": 0.1061,
+      "step": 9167
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 8.791352331191855e-06,
+      "loss": 0.0959,
+      "step": 9168
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 8.791219044064724e-06,
+      "loss": 0.0964,
+      "step": 9169
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 8.791085756937595e-06,
+      "loss": 0.1425,
+      "step": 9170
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 8.790952469810466e-06,
+      "loss": 0.0867,
+      "step": 9171
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 8.790819182683337e-06,
+      "loss": 0.0647,
+      "step": 9172
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 8.790685895556208e-06,
+      "loss": 0.1601,
+      "step": 9173
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 8.790552608429077e-06,
+      "loss": 0.0774,
+      "step": 9174
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 8.79041932130195e-06,
+      "loss": 0.1207,
+      "step": 9175
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 8.79028603417482e-06,
+      "loss": 0.0998,
+      "step": 9176
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 8.79015274704769e-06,
+      "loss": 0.0981,
+      "step": 9177
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 8.790019459920561e-06,
+      "loss": 0.0774,
+      "step": 9178
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 8.789886172793432e-06,
+      "loss": 0.0885,
+      "step": 9179
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 8.789752885666303e-06,
+      "loss": 0.0917,
+      "step": 9180
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 8.789619598539174e-06,
+      "loss": 0.0736,
+      "step": 9181
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 8.789486311412044e-06,
+      "loss": 0.0922,
+      "step": 9182
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 8.789353024284916e-06,
+      "loss": 0.1657,
+      "step": 9183
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 8.789219737157786e-06,
+      "loss": 0.1375,
+      "step": 9184
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 8.789086450030657e-06,
+      "loss": 0.0967,
+      "step": 9185
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 8.788953162903528e-06,
+      "loss": 0.0876,
+      "step": 9186
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 8.788819875776399e-06,
+      "loss": 0.1363,
+      "step": 9187
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 8.78868658864927e-06,
+      "loss": 0.137,
+      "step": 9188
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 8.78855330152214e-06,
+      "loss": 0.0975,
+      "step": 9189
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 8.78842001439501e-06,
+      "loss": 0.0879,
+      "step": 9190
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 8.788286727267881e-06,
+      "loss": 0.1337,
+      "step": 9191
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 8.788153440140752e-06,
+      "loss": 0.1354,
+      "step": 9192
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 8.788020153013623e-06,
+      "loss": 0.1271,
+      "step": 9193
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 8.787886865886494e-06,
+      "loss": 0.0515,
+      "step": 9194
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 8.787753578759363e-06,
+      "loss": 0.1101,
+      "step": 9195
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 8.787620291632234e-06,
+      "loss": 0.0608,
+      "step": 9196
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 8.787487004505105e-06,
+      "loss": 0.0942,
+      "step": 9197
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 8.787353717377976e-06,
+      "loss": 0.1413,
+      "step": 9198
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 8.787220430250847e-06,
+      "loss": 0.1068,
+      "step": 9199
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 8.787087143123718e-06,
+      "loss": 0.1142,
+      "step": 9200
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 8.786953855996588e-06,
+      "loss": 0.1374,
+      "step": 9201
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 8.78682056886946e-06,
+      "loss": 0.1107,
+      "step": 9202
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 8.78668728174233e-06,
+      "loss": 0.1237,
+      "step": 9203
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 8.7865539946152e-06,
+      "loss": 0.085,
+      "step": 9204
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 8.786420707488072e-06,
+      "loss": 0.1196,
+      "step": 9205
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 8.786287420360942e-06,
+      "loss": 0.1139,
+      "step": 9206
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 8.786154133233813e-06,
+      "loss": 0.0958,
+      "step": 9207
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 8.786020846106684e-06,
+      "loss": 0.0879,
+      "step": 9208
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 8.785887558979554e-06,
+      "loss": 0.1078,
+      "step": 9209
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 8.785754271852425e-06,
+      "loss": 0.1221,
+      "step": 9210
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 8.785620984725296e-06,
+      "loss": 0.0896,
+      "step": 9211
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 8.785487697598167e-06,
+      "loss": 0.0954,
+      "step": 9212
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 8.785354410471038e-06,
+      "loss": 0.0935,
+      "step": 9213
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 8.785221123343907e-06,
+      "loss": 0.1451,
+      "step": 9214
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 8.78508783621678e-06,
+      "loss": 0.1116,
+      "step": 9215
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 8.784954549089649e-06,
+      "loss": 0.1011,
+      "step": 9216
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 8.78482126196252e-06,
+      "loss": 0.1111,
+      "step": 9217
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 8.784687974835391e-06,
+      "loss": 0.1152,
+      "step": 9218
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 8.784554687708262e-06,
+      "loss": 0.1224,
+      "step": 9219
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 8.784421400581133e-06,
+      "loss": 0.0795,
+      "step": 9220
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 8.784288113454004e-06,
+      "loss": 0.102,
+      "step": 9221
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 8.784154826326873e-06,
+      "loss": 0.1034,
+      "step": 9222
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 8.784021539199746e-06,
+      "loss": 0.1291,
+      "step": 9223
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 8.783888252072615e-06,
+      "loss": 0.0927,
+      "step": 9224
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 8.783754964945486e-06,
+      "loss": 0.0906,
+      "step": 9225
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 8.783621677818357e-06,
+      "loss": 0.105,
+      "step": 9226
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 8.783488390691228e-06,
+      "loss": 0.0844,
+      "step": 9227
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 8.7833551035641e-06,
+      "loss": 0.1082,
+      "step": 9228
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 8.783221816436969e-06,
+      "loss": 0.123,
+      "step": 9229
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 8.78308852930984e-06,
+      "loss": 0.0808,
+      "step": 9230
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 8.78295524218271e-06,
+      "loss": 0.119,
+      "step": 9231
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 8.782821955055582e-06,
+      "loss": 0.1079,
+      "step": 9232
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 8.782688667928453e-06,
+      "loss": 0.0931,
+      "step": 9233
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 8.782555380801324e-06,
+      "loss": 0.1467,
+      "step": 9234
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 8.782422093674193e-06,
+      "loss": 0.1054,
+      "step": 9235
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 8.782288806547064e-06,
+      "loss": 0.1363,
+      "step": 9236
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 8.782155519419935e-06,
+      "loss": 0.1213,
+      "step": 9237
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 8.782022232292806e-06,
+      "loss": 0.07,
+      "step": 9238
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 8.781888945165677e-06,
+      "loss": 0.1213,
+      "step": 9239
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 8.781755658038548e-06,
+      "loss": 0.0896,
+      "step": 9240
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 8.781622370911417e-06,
+      "loss": 0.0881,
+      "step": 9241
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 8.78148908378429e-06,
+      "loss": 0.0945,
+      "step": 9242
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 8.781355796657159e-06,
+      "loss": 0.1247,
+      "step": 9243
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 8.78122250953003e-06,
+      "loss": 0.0799,
+      "step": 9244
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 8.781089222402901e-06,
+      "loss": 0.0756,
+      "step": 9245
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 8.780955935275772e-06,
+      "loss": 0.1286,
+      "step": 9246
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 8.780822648148643e-06,
+      "loss": 0.1162,
+      "step": 9247
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 8.780689361021514e-06,
+      "loss": 0.1025,
+      "step": 9248
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 8.780556073894383e-06,
+      "loss": 0.0748,
+      "step": 9249
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 8.780422786767254e-06,
+      "loss": 0.1149,
+      "step": 9250
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 8.780289499640125e-06,
+      "loss": 0.091,
+      "step": 9251
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 8.780156212512996e-06,
+      "loss": 0.1234,
+      "step": 9252
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 8.780022925385867e-06,
+      "loss": 0.1579,
+      "step": 9253
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 8.779889638258737e-06,
+      "loss": 0.1362,
+      "step": 9254
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 8.77975635113161e-06,
+      "loss": 0.1151,
+      "step": 9255
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 8.779623064004479e-06,
+      "loss": 0.0911,
+      "step": 9256
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 8.77948977687735e-06,
+      "loss": 0.0938,
+      "step": 9257
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 8.77935648975022e-06,
+      "loss": 0.0564,
+      "step": 9258
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 8.779223202623092e-06,
+      "loss": 0.1055,
+      "step": 9259
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 8.779089915495963e-06,
+      "loss": 0.0867,
+      "step": 9260
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 8.778956628368834e-06,
+      "loss": 0.0899,
+      "step": 9261
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 8.778823341241703e-06,
+      "loss": 0.0998,
+      "step": 9262
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 8.778690054114574e-06,
+      "loss": 0.158,
+      "step": 9263
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 8.778556766987445e-06,
+      "loss": 0.0845,
+      "step": 9264
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 8.778423479860316e-06,
+      "loss": 0.074,
+      "step": 9265
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 8.778290192733187e-06,
+      "loss": 0.0843,
+      "step": 9266
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 8.778156905606058e-06,
+      "loss": 0.0884,
+      "step": 9267
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 8.778023618478927e-06,
+      "loss": 0.1424,
+      "step": 9268
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 8.777890331351798e-06,
+      "loss": 0.1317,
+      "step": 9269
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 8.77775704422467e-06,
+      "loss": 0.1073,
+      "step": 9270
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 8.77762375709754e-06,
+      "loss": 0.0986,
+      "step": 9271
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 8.777490469970411e-06,
+      "loss": 0.1461,
+      "step": 9272
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 8.77735718284328e-06,
+      "loss": 0.1108,
+      "step": 9273
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 8.777223895716153e-06,
+      "loss": 0.1414,
+      "step": 9274
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 8.777090608589022e-06,
+      "loss": 0.1177,
+      "step": 9275
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 8.776957321461893e-06,
+      "loss": 0.114,
+      "step": 9276
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 8.776824034334764e-06,
+      "loss": 0.0613,
+      "step": 9277
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 8.776690747207635e-06,
+      "loss": 0.1194,
+      "step": 9278
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 8.776557460080506e-06,
+      "loss": 0.0844,
+      "step": 9279
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 8.776424172953377e-06,
+      "loss": 0.1244,
+      "step": 9280
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 8.776290885826247e-06,
+      "loss": 0.1054,
+      "step": 9281
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 8.77615759869912e-06,
+      "loss": 0.1675,
+      "step": 9282
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 8.776024311571989e-06,
+      "loss": 0.1111,
+      "step": 9283
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 8.77589102444486e-06,
+      "loss": 0.1364,
+      "step": 9284
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 8.77575773731773e-06,
+      "loss": 0.1033,
+      "step": 9285
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 8.775624450190602e-06,
+      "loss": 0.0873,
+      "step": 9286
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 8.775491163063473e-06,
+      "loss": 0.0701,
+      "step": 9287
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 8.775357875936342e-06,
+      "loss": 0.135,
+      "step": 9288
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 8.775224588809213e-06,
+      "loss": 0.1098,
+      "step": 9289
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 8.775091301682084e-06,
+      "loss": 0.0983,
+      "step": 9290
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 8.774958014554955e-06,
+      "loss": 0.0984,
+      "step": 9291
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 8.774824727427826e-06,
+      "loss": 0.1011,
+      "step": 9292
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 8.774691440300697e-06,
+      "loss": 0.1019,
+      "step": 9293
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 8.774558153173566e-06,
+      "loss": 0.1284,
+      "step": 9294
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 8.774424866046439e-06,
+      "loss": 0.0899,
+      "step": 9295
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 8.774291578919308e-06,
+      "loss": 0.1384,
+      "step": 9296
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 8.77415829179218e-06,
+      "loss": 0.1774,
+      "step": 9297
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 8.77402500466505e-06,
+      "loss": 0.1149,
+      "step": 9298
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 8.773891717537921e-06,
+      "loss": 0.0774,
+      "step": 9299
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 8.773758430410792e-06,
+      "loss": 0.0822,
+      "step": 9300
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 8.773625143283663e-06,
+      "loss": 0.0938,
+      "step": 9301
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 8.773491856156533e-06,
+      "loss": 0.0854,
+      "step": 9302
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 8.773358569029404e-06,
+      "loss": 0.1675,
+      "step": 9303
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 8.773225281902275e-06,
+      "loss": 0.0917,
+      "step": 9304
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 8.773091994775146e-06,
+      "loss": 0.095,
+      "step": 9305
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 8.772958707648017e-06,
+      "loss": 0.0923,
+      "step": 9306
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 8.772825420520886e-06,
+      "loss": 0.0776,
+      "step": 9307
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 8.772692133393757e-06,
+      "loss": 0.1041,
+      "step": 9308
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 8.772558846266628e-06,
+      "loss": 0.1029,
+      "step": 9309
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 8.772425559139499e-06,
+      "loss": 0.0721,
+      "step": 9310
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 8.77229227201237e-06,
+      "loss": 0.1147,
+      "step": 9311
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 8.77215898488524e-06,
+      "loss": 0.0982,
+      "step": 9312
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 8.77202569775811e-06,
+      "loss": 0.1133,
+      "step": 9313
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 8.771892410630983e-06,
+      "loss": 0.1052,
+      "step": 9314
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 8.771759123503852e-06,
+      "loss": 0.0756,
+      "step": 9315
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 8.771625836376723e-06,
+      "loss": 0.0955,
+      "step": 9316
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 8.771492549249594e-06,
+      "loss": 0.0723,
+      "step": 9317
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 8.771359262122465e-06,
+      "loss": 0.0732,
+      "step": 9318
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 8.771225974995336e-06,
+      "loss": 0.0891,
+      "step": 9319
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 8.771092687868207e-06,
+      "loss": 0.0752,
+      "step": 9320
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 8.770959400741076e-06,
+      "loss": 0.1364,
+      "step": 9321
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 8.770826113613949e-06,
+      "loss": 0.0941,
+      "step": 9322
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 8.770692826486818e-06,
+      "loss": 0.0986,
+      "step": 9323
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 8.77055953935969e-06,
+      "loss": 0.0832,
+      "step": 9324
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 8.77042625223256e-06,
+      "loss": 0.1242,
+      "step": 9325
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 8.770292965105431e-06,
+      "loss": 0.0979,
+      "step": 9326
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 8.770159677978302e-06,
+      "loss": 0.1061,
+      "step": 9327
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 8.770026390851172e-06,
+      "loss": 0.1085,
+      "step": 9328
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 8.769893103724043e-06,
+      "loss": 0.1172,
+      "step": 9329
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 8.769759816596914e-06,
+      "loss": 0.1183,
+      "step": 9330
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 8.769626529469785e-06,
+      "loss": 0.0945,
+      "step": 9331
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 8.769493242342656e-06,
+      "loss": 0.0717,
+      "step": 9332
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 8.769359955215527e-06,
+      "loss": 0.103,
+      "step": 9333
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 8.769226668088396e-06,
+      "loss": 0.0617,
+      "step": 9334
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 8.769093380961267e-06,
+      "loss": 0.0493,
+      "step": 9335
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 8.768960093834138e-06,
+      "loss": 0.0701,
+      "step": 9336
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 8.768826806707009e-06,
+      "loss": 0.0995,
+      "step": 9337
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 8.76869351957988e-06,
+      "loss": 0.0917,
+      "step": 9338
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 8.768560232452751e-06,
+      "loss": 0.1238,
+      "step": 9339
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 8.76842694532562e-06,
+      "loss": 0.0934,
+      "step": 9340
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 8.768293658198493e-06,
+      "loss": 0.1022,
+      "step": 9341
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 8.768160371071362e-06,
+      "loss": 0.0531,
+      "step": 9342
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 8.768027083944233e-06,
+      "loss": 0.1018,
+      "step": 9343
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 8.767893796817104e-06,
+      "loss": 0.125,
+      "step": 9344
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 8.767760509689975e-06,
+      "loss": 0.1209,
+      "step": 9345
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 8.767627222562846e-06,
+      "loss": 0.1389,
+      "step": 9346
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 8.767493935435715e-06,
+      "loss": 0.1177,
+      "step": 9347
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 8.767360648308586e-06,
+      "loss": 0.1114,
+      "step": 9348
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 8.767227361181457e-06,
+      "loss": 0.0983,
+      "step": 9349
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 8.767094074054328e-06,
+      "loss": 0.0977,
+      "step": 9350
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 8.7669607869272e-06,
+      "loss": 0.0937,
+      "step": 9351
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 8.76682749980007e-06,
+      "loss": 0.1033,
+      "step": 9352
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 8.76669421267294e-06,
+      "loss": 0.1112,
+      "step": 9353
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 8.766560925545812e-06,
+      "loss": 0.102,
+      "step": 9354
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 8.766427638418682e-06,
+      "loss": 0.1142,
+      "step": 9355
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 8.766294351291553e-06,
+      "loss": 0.1065,
+      "step": 9356
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 8.766161064164424e-06,
+      "loss": 0.0774,
+      "step": 9357
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 8.766027777037295e-06,
+      "loss": 0.1729,
+      "step": 9358
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 8.765894489910166e-06,
+      "loss": 0.0779,
+      "step": 9359
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 8.765761202783037e-06,
+      "loss": 0.1057,
+      "step": 9360
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 8.765627915655906e-06,
+      "loss": 0.0906,
+      "step": 9361
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 8.765494628528777e-06,
+      "loss": 0.0891,
+      "step": 9362
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 8.765361341401648e-06,
+      "loss": 0.1101,
+      "step": 9363
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 8.765228054274519e-06,
+      "loss": 0.0992,
+      "step": 9364
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 8.76509476714739e-06,
+      "loss": 0.1175,
+      "step": 9365
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 8.76496148002026e-06,
+      "loss": 0.13,
+      "step": 9366
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 8.76482819289313e-06,
+      "loss": 0.1039,
+      "step": 9367
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 8.764694905766001e-06,
+      "loss": 0.0704,
+      "step": 9368
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 8.764561618638872e-06,
+      "loss": 0.1213,
+      "step": 9369
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 8.764428331511743e-06,
+      "loss": 0.1184,
+      "step": 9370
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 8.764295044384614e-06,
+      "loss": 0.0961,
+      "step": 9371
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 8.764161757257484e-06,
+      "loss": 0.0813,
+      "step": 9372
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 8.764028470130356e-06,
+      "loss": 0.137,
+      "step": 9373
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 8.763895183003226e-06,
+      "loss": 0.1167,
+      "step": 9374
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 8.763761895876097e-06,
+      "loss": 0.0885,
+      "step": 9375
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 8.763628608748968e-06,
+      "loss": 0.1021,
+      "step": 9376
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 8.763495321621839e-06,
+      "loss": 0.1434,
+      "step": 9377
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 8.76336203449471e-06,
+      "loss": 0.0943,
+      "step": 9378
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 8.76322874736758e-06,
+      "loss": 0.1272,
+      "step": 9379
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 8.76309546024045e-06,
+      "loss": 0.1162,
+      "step": 9380
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 8.762962173113323e-06,
+      "loss": 0.0776,
+      "step": 9381
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 8.762828885986192e-06,
+      "loss": 0.1097,
+      "step": 9382
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 8.762695598859063e-06,
+      "loss": 0.1145,
+      "step": 9383
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 8.762562311731934e-06,
+      "loss": 0.1089,
+      "step": 9384
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 8.762429024604805e-06,
+      "loss": 0.1276,
+      "step": 9385
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 8.762295737477676e-06,
+      "loss": 0.1078,
+      "step": 9386
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 8.762162450350545e-06,
+      "loss": 0.0951,
+      "step": 9387
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 8.762029163223416e-06,
+      "loss": 0.1148,
+      "step": 9388
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 8.761895876096287e-06,
+      "loss": 0.1046,
+      "step": 9389
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 8.761762588969158e-06,
+      "loss": 0.0848,
+      "step": 9390
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 8.761629301842029e-06,
+      "loss": 0.0909,
+      "step": 9391
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 8.7614960147149e-06,
+      "loss": 0.0958,
+      "step": 9392
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 8.76136272758777e-06,
+      "loss": 0.1059,
+      "step": 9393
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 8.761229440460642e-06,
+      "loss": 0.127,
+      "step": 9394
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 8.761096153333511e-06,
+      "loss": 0.1313,
+      "step": 9395
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 8.760962866206382e-06,
+      "loss": 0.08,
+      "step": 9396
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 8.760829579079253e-06,
+      "loss": 0.1617,
+      "step": 9397
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 8.760696291952124e-06,
+      "loss": 0.1215,
+      "step": 9398
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 8.760563004824995e-06,
+      "loss": 0.1212,
+      "step": 9399
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 8.760429717697866e-06,
+      "loss": 0.1035,
+      "step": 9400
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 8.760296430570736e-06,
+      "loss": 0.1065,
+      "step": 9401
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 8.760163143443607e-06,
+      "loss": 0.0902,
+      "step": 9402
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 8.760029856316478e-06,
+      "loss": 0.0891,
+      "step": 9403
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 8.759896569189349e-06,
+      "loss": 0.1046,
+      "step": 9404
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 8.75976328206222e-06,
+      "loss": 0.0602,
+      "step": 9405
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 8.759629994935089e-06,
+      "loss": 0.129,
+      "step": 9406
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 8.75949670780796e-06,
+      "loss": 0.1172,
+      "step": 9407
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 8.759363420680831e-06,
+      "loss": 0.1108,
+      "step": 9408
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 8.759230133553702e-06,
+      "loss": 0.1117,
+      "step": 9409
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 8.759096846426573e-06,
+      "loss": 0.0942,
+      "step": 9410
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 8.758963559299444e-06,
+      "loss": 0.0925,
+      "step": 9411
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 8.758830272172313e-06,
+      "loss": 0.1047,
+      "step": 9412
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 8.758696985045186e-06,
+      "loss": 0.0778,
+      "step": 9413
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 8.758563697918055e-06,
+      "loss": 0.1305,
+      "step": 9414
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 8.758430410790926e-06,
+      "loss": 0.1122,
+      "step": 9415
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 8.758297123663797e-06,
+      "loss": 0.0836,
+      "step": 9416
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 8.758163836536668e-06,
+      "loss": 0.0947,
+      "step": 9417
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 8.75803054940954e-06,
+      "loss": 0.12,
+      "step": 9418
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 8.75789726228241e-06,
+      "loss": 0.0691,
+      "step": 9419
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 8.75776397515528e-06,
+      "loss": 0.0894,
+      "step": 9420
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 8.757630688028152e-06,
+      "loss": 0.1081,
+      "step": 9421
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 8.757497400901021e-06,
+      "loss": 0.0864,
+      "step": 9422
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 8.757364113773892e-06,
+      "loss": 0.1112,
+      "step": 9423
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 8.757230826646763e-06,
+      "loss": 0.1039,
+      "step": 9424
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 8.757097539519633e-06,
+      "loss": 0.0999,
+      "step": 9425
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 8.756964252392505e-06,
+      "loss": 0.1477,
+      "step": 9426
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 8.756830965265375e-06,
+      "loss": 0.124,
+      "step": 9427
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 8.756697678138246e-06,
+      "loss": 0.1074,
+      "step": 9428
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 8.756564391011117e-06,
+      "loss": 0.0507,
+      "step": 9429
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 8.756431103883988e-06,
+      "loss": 0.1145,
+      "step": 9430
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 8.756297816756859e-06,
+      "loss": 0.1071,
+      "step": 9431
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 8.75616452962973e-06,
+      "loss": 0.1052,
+      "step": 9432
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 8.756031242502599e-06,
+      "loss": 0.1055,
+      "step": 9433
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 8.75589795537547e-06,
+      "loss": 0.1151,
+      "step": 9434
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 8.755764668248341e-06,
+      "loss": 0.1045,
+      "step": 9435
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 8.755631381121212e-06,
+      "loss": 0.1158,
+      "step": 9436
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 8.755498093994083e-06,
+      "loss": 0.106,
+      "step": 9437
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 8.755364806866954e-06,
+      "loss": 0.073,
+      "step": 9438
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 8.755231519739823e-06,
+      "loss": 0.1134,
+      "step": 9439
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 8.755098232612696e-06,
+      "loss": 0.0937,
+      "step": 9440
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 8.754964945485565e-06,
+      "loss": 0.0949,
+      "step": 9441
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 8.754831658358436e-06,
+      "loss": 0.0881,
+      "step": 9442
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 8.754698371231307e-06,
+      "loss": 0.1508,
+      "step": 9443
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 8.754565084104177e-06,
+      "loss": 0.1213,
+      "step": 9444
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 8.75443179697705e-06,
+      "loss": 0.0899,
+      "step": 9445
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 8.754298509849919e-06,
+      "loss": 0.102,
+      "step": 9446
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 8.75416522272279e-06,
+      "loss": 0.1496,
+      "step": 9447
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 8.75403193559566e-06,
+      "loss": 0.139,
+      "step": 9448
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 8.753898648468532e-06,
+      "loss": 0.1085,
+      "step": 9449
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 8.753765361341403e-06,
+      "loss": 0.079,
+      "step": 9450
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 8.753632074214274e-06,
+      "loss": 0.0723,
+      "step": 9451
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 8.753498787087143e-06,
+      "loss": 0.114,
+      "step": 9452
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 8.753365499960016e-06,
+      "loss": 0.0844,
+      "step": 9453
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 8.753232212832885e-06,
+      "loss": 0.1017,
+      "step": 9454
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 8.753098925705756e-06,
+      "loss": 0.085,
+      "step": 9455
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 8.752965638578627e-06,
+      "loss": 0.0893,
+      "step": 9456
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 8.752832351451498e-06,
+      "loss": 0.0637,
+      "step": 9457
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 8.752699064324369e-06,
+      "loss": 0.1008,
+      "step": 9458
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 8.75256577719724e-06,
+      "loss": 0.1057,
+      "step": 9459
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 8.752432490070109e-06,
+      "loss": 0.1193,
+      "step": 9460
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 8.752299202942982e-06,
+      "loss": 0.1228,
+      "step": 9461
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 8.752165915815851e-06,
+      "loss": 0.0598,
+      "step": 9462
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 8.752032628688722e-06,
+      "loss": 0.1007,
+      "step": 9463
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 8.751899341561593e-06,
+      "loss": 0.109,
+      "step": 9464
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 8.751766054434462e-06,
+      "loss": 0.0907,
+      "step": 9465
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 8.751632767307335e-06,
+      "loss": 0.0961,
+      "step": 9466
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 8.751499480180204e-06,
+      "loss": 0.0794,
+      "step": 9467
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 8.751366193053075e-06,
+      "loss": 0.1138,
+      "step": 9468
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 8.751232905925946e-06,
+      "loss": 0.0702,
+      "step": 9469
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 8.751099618798817e-06,
+      "loss": 0.1621,
+      "step": 9470
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 8.750966331671688e-06,
+      "loss": 0.111,
+      "step": 9471
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 8.75083304454456e-06,
+      "loss": 0.1049,
+      "step": 9472
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 8.750699757417429e-06,
+      "loss": 0.0929,
+      "step": 9473
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 8.7505664702903e-06,
+      "loss": 0.068,
+      "step": 9474
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 8.75043318316317e-06,
+      "loss": 0.0857,
+      "step": 9475
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 8.750299896036042e-06,
+      "loss": 0.1099,
+      "step": 9476
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 8.750166608908913e-06,
+      "loss": 0.1149,
+      "step": 9477
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 8.750033321781784e-06,
+      "loss": 0.1318,
+      "step": 9478
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 8.749900034654653e-06,
+      "loss": 0.0782,
+      "step": 9479
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 8.749766747527526e-06,
+      "loss": 0.0949,
+      "step": 9480
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 8.749633460400395e-06,
+      "loss": 0.066,
+      "step": 9481
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 8.749500173273266e-06,
+      "loss": 0.1007,
+      "step": 9482
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 8.749366886146137e-06,
+      "loss": 0.1215,
+      "step": 9483
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 8.749233599019006e-06,
+      "loss": 0.1169,
+      "step": 9484
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 8.749100311891879e-06,
+      "loss": 0.1084,
+      "step": 9485
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 8.748967024764748e-06,
+      "loss": 0.1179,
+      "step": 9486
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 8.74883373763762e-06,
+      "loss": 0.0692,
+      "step": 9487
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 8.74870045051049e-06,
+      "loss": 0.1231,
+      "step": 9488
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 8.748567163383361e-06,
+      "loss": 0.1124,
+      "step": 9489
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 8.748433876256232e-06,
+      "loss": 0.0976,
+      "step": 9490
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 8.748300589129103e-06,
+      "loss": 0.1254,
+      "step": 9491
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 8.748167302001972e-06,
+      "loss": 0.1244,
+      "step": 9492
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 8.748034014874845e-06,
+      "loss": 0.1084,
+      "step": 9493
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 8.747900727747714e-06,
+      "loss": 0.1315,
+      "step": 9494
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 8.747767440620585e-06,
+      "loss": 0.0783,
+      "step": 9495
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 8.747634153493456e-06,
+      "loss": 0.0793,
+      "step": 9496
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 8.747500866366327e-06,
+      "loss": 0.1206,
+      "step": 9497
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 8.747367579239198e-06,
+      "loss": 0.1117,
+      "step": 9498
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 8.74723429211207e-06,
+      "loss": 0.0884,
+      "step": 9499
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 8.747101004984939e-06,
+      "loss": 0.1176,
+      "step": 9500
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 8.74696771785781e-06,
+      "loss": 0.1063,
+      "step": 9501
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 8.74683443073068e-06,
+      "loss": 0.1357,
+      "step": 9502
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 8.746701143603552e-06,
+      "loss": 0.123,
+      "step": 9503
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 8.746567856476423e-06,
+      "loss": 0.101,
+      "step": 9504
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 8.746434569349292e-06,
+      "loss": 0.0894,
+      "step": 9505
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 8.746301282222163e-06,
+      "loss": 0.1014,
+      "step": 9506
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 8.746167995095034e-06,
+      "loss": 0.1223,
+      "step": 9507
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 8.746034707967905e-06,
+      "loss": 0.0984,
+      "step": 9508
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 8.745901420840776e-06,
+      "loss": 0.1232,
+      "step": 9509
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 8.745768133713647e-06,
+      "loss": 0.1012,
+      "step": 9510
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 8.745634846586516e-06,
+      "loss": 0.09,
+      "step": 9511
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 8.745501559459389e-06,
+      "loss": 0.1123,
+      "step": 9512
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 8.745368272332258e-06,
+      "loss": 0.111,
+      "step": 9513
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 8.74523498520513e-06,
+      "loss": 0.0774,
+      "step": 9514
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 8.745101698078e-06,
+      "loss": 0.1303,
+      "step": 9515
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 8.744968410950871e-06,
+      "loss": 0.0944,
+      "step": 9516
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 8.744835123823742e-06,
+      "loss": 0.0756,
+      "step": 9517
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 8.744701836696613e-06,
+      "loss": 0.0955,
+      "step": 9518
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 8.744568549569483e-06,
+      "loss": 0.1415,
+      "step": 9519
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 8.744435262442355e-06,
+      "loss": 0.1197,
+      "step": 9520
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 8.744301975315225e-06,
+      "loss": 0.0813,
+      "step": 9521
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 8.744168688188096e-06,
+      "loss": 0.1037,
+      "step": 9522
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 8.744035401060967e-06,
+      "loss": 0.12,
+      "step": 9523
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 8.743902113933836e-06,
+      "loss": 0.1018,
+      "step": 9524
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 8.743768826806708e-06,
+      "loss": 0.107,
+      "step": 9525
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 8.743635539679578e-06,
+      "loss": 0.1003,
+      "step": 9526
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 8.743502252552449e-06,
+      "loss": 0.0837,
+      "step": 9527
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 8.74336896542532e-06,
+      "loss": 0.0982,
+      "step": 9528
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 8.74323567829819e-06,
+      "loss": 0.1411,
+      "step": 9529
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 8.743102391171062e-06,
+      "loss": 0.1012,
+      "step": 9530
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 8.742969104043933e-06,
+      "loss": 0.1464,
+      "step": 9531
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 8.742835816916802e-06,
+      "loss": 0.109,
+      "step": 9532
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 8.742702529789675e-06,
+      "loss": 0.1137,
+      "step": 9533
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 8.742569242662544e-06,
+      "loss": 0.0888,
+      "step": 9534
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 8.742435955535415e-06,
+      "loss": 0.1009,
+      "step": 9535
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 8.742302668408286e-06,
+      "loss": 0.1142,
+      "step": 9536
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 8.742169381281157e-06,
+      "loss": 0.1187,
+      "step": 9537
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 8.742036094154028e-06,
+      "loss": 0.1341,
+      "step": 9538
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 8.741902807026899e-06,
+      "loss": 0.11,
+      "step": 9539
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 8.741769519899768e-06,
+      "loss": 0.0839,
+      "step": 9540
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 8.74163623277264e-06,
+      "loss": 0.1194,
+      "step": 9541
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 8.74150294564551e-06,
+      "loss": 0.1259,
+      "step": 9542
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 8.741369658518381e-06,
+      "loss": 0.1034,
+      "step": 9543
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 8.741236371391252e-06,
+      "loss": 0.1071,
+      "step": 9544
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 8.741103084264122e-06,
+      "loss": 0.0957,
+      "step": 9545
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 8.740969797136993e-06,
+      "loss": 0.0828,
+      "step": 9546
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 8.740836510009864e-06,
+      "loss": 0.102,
+      "step": 9547
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 8.740703222882735e-06,
+      "loss": 0.1137,
+      "step": 9548
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 8.740569935755606e-06,
+      "loss": 0.1539,
+      "step": 9549
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 8.740436648628477e-06,
+      "loss": 0.1183,
+      "step": 9550
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 8.740303361501346e-06,
+      "loss": 0.0939,
+      "step": 9551
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 8.740170074374219e-06,
+      "loss": 0.1102,
+      "step": 9552
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 8.740036787247088e-06,
+      "loss": 0.0699,
+      "step": 9553
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 8.739903500119959e-06,
+      "loss": 0.1299,
+      "step": 9554
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 8.73977021299283e-06,
+      "loss": 0.109,
+      "step": 9555
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 8.739636925865701e-06,
+      "loss": 0.1076,
+      "step": 9556
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 8.739503638738572e-06,
+      "loss": 0.1021,
+      "step": 9557
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 8.739370351611443e-06,
+      "loss": 0.1299,
+      "step": 9558
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 8.739237064484312e-06,
+      "loss": 0.109,
+      "step": 9559
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 8.739103777357185e-06,
+      "loss": 0.1049,
+      "step": 9560
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 8.738970490230054e-06,
+      "loss": 0.139,
+      "step": 9561
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 8.738837203102925e-06,
+      "loss": 0.1425,
+      "step": 9562
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 8.738703915975796e-06,
+      "loss": 0.0942,
+      "step": 9563
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 8.738570628848665e-06,
+      "loss": 0.1341,
+      "step": 9564
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 8.738437341721538e-06,
+      "loss": 0.1245,
+      "step": 9565
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 8.738304054594407e-06,
+      "loss": 0.1531,
+      "step": 9566
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 8.738170767467278e-06,
+      "loss": 0.082,
+      "step": 9567
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 8.73803748034015e-06,
+      "loss": 0.0658,
+      "step": 9568
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 8.73790419321302e-06,
+      "loss": 0.1094,
+      "step": 9569
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 8.737770906085891e-06,
+      "loss": 0.0684,
+      "step": 9570
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 8.737637618958762e-06,
+      "loss": 0.1251,
+      "step": 9571
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 8.737504331831632e-06,
+      "loss": 0.1062,
+      "step": 9572
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 8.737371044704503e-06,
+      "loss": 0.0834,
+      "step": 9573
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 8.737237757577374e-06,
+      "loss": 0.0921,
+      "step": 9574
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 8.737104470450245e-06,
+      "loss": 0.1288,
+      "step": 9575
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 8.736971183323116e-06,
+      "loss": 0.0682,
+      "step": 9576
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 8.736837896195987e-06,
+      "loss": 0.1013,
+      "step": 9577
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 8.736704609068856e-06,
+      "loss": 0.0984,
+      "step": 9578
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 8.736571321941729e-06,
+      "loss": 0.1737,
+      "step": 9579
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 8.736438034814598e-06,
+      "loss": 0.1315,
+      "step": 9580
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 8.736304747687469e-06,
+      "loss": 0.117,
+      "step": 9581
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 8.73617146056034e-06,
+      "loss": 0.1234,
+      "step": 9582
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 8.73603817343321e-06,
+      "loss": 0.0643,
+      "step": 9583
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 8.735904886306082e-06,
+      "loss": 0.1,
+      "step": 9584
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 8.735771599178951e-06,
+      "loss": 0.0894,
+      "step": 9585
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 8.735638312051822e-06,
+      "loss": 0.1354,
+      "step": 9586
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 8.735505024924693e-06,
+      "loss": 0.0887,
+      "step": 9587
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 8.735371737797564e-06,
+      "loss": 0.1595,
+      "step": 9588
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 8.735238450670435e-06,
+      "loss": 0.1223,
+      "step": 9589
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 8.735105163543306e-06,
+      "loss": 0.136,
+      "step": 9590
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 8.734971876416176e-06,
+      "loss": 0.0977,
+      "step": 9591
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 8.734838589289048e-06,
+      "loss": 0.0986,
+      "step": 9592
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 8.734705302161918e-06,
+      "loss": 0.1218,
+      "step": 9593
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 8.734572015034788e-06,
+      "loss": 0.1268,
+      "step": 9594
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 8.73443872790766e-06,
+      "loss": 0.1211,
+      "step": 9595
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 8.73430544078053e-06,
+      "loss": 0.0999,
+      "step": 9596
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 8.734172153653401e-06,
+      "loss": 0.0981,
+      "step": 9597
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 8.734038866526272e-06,
+      "loss": 0.0991,
+      "step": 9598
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 8.733905579399142e-06,
+      "loss": 0.1715,
+      "step": 9599
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 8.733772292272014e-06,
+      "loss": 0.1184,
+      "step": 9600
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 8.733639005144884e-06,
+      "loss": 0.09,
+      "step": 9601
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 8.733505718017755e-06,
+      "loss": 0.11,
+      "step": 9602
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 8.733372430890626e-06,
+      "loss": 0.098,
+      "step": 9603
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 8.733239143763495e-06,
+      "loss": 0.1336,
+      "step": 9604
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 8.733105856636368e-06,
+      "loss": 0.068,
+      "step": 9605
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 8.732972569509237e-06,
+      "loss": 0.1049,
+      "step": 9606
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 8.732839282382108e-06,
+      "loss": 0.1269,
+      "step": 9607
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 8.732705995254979e-06,
+      "loss": 0.0975,
+      "step": 9608
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 8.73257270812785e-06,
+      "loss": 0.1145,
+      "step": 9609
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 8.73243942100072e-06,
+      "loss": 0.0808,
+      "step": 9610
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 8.732306133873592e-06,
+      "loss": 0.1259,
+      "step": 9611
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 8.732172846746461e-06,
+      "loss": 0.1065,
+      "step": 9612
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 8.732039559619332e-06,
+      "loss": 0.0986,
+      "step": 9613
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 8.731906272492203e-06,
+      "loss": 0.1305,
+      "step": 9614
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 8.731772985365074e-06,
+      "loss": 0.1137,
+      "step": 9615
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 8.731639698237945e-06,
+      "loss": 0.1128,
+      "step": 9616
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 8.731506411110816e-06,
+      "loss": 0.1015,
+      "step": 9617
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 8.731373123983686e-06,
+      "loss": 0.0984,
+      "step": 9618
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 8.731239836856558e-06,
+      "loss": 0.0897,
+      "step": 9619
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 8.731106549729428e-06,
+      "loss": 0.1044,
+      "step": 9620
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 8.730973262602299e-06,
+      "loss": 0.0894,
+      "step": 9621
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 8.73083997547517e-06,
+      "loss": 0.1224,
+      "step": 9622
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 8.730706688348039e-06,
+      "loss": 0.1272,
+      "step": 9623
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 8.730573401220912e-06,
+      "loss": 0.1047,
+      "step": 9624
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 8.730440114093781e-06,
+      "loss": 0.1916,
+      "step": 9625
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 8.730306826966652e-06,
+      "loss": 0.0922,
+      "step": 9626
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 8.730173539839523e-06,
+      "loss": 0.1049,
+      "step": 9627
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 8.730040252712394e-06,
+      "loss": 0.0867,
+      "step": 9628
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 8.729906965585265e-06,
+      "loss": 0.1446,
+      "step": 9629
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 8.729773678458136e-06,
+      "loss": 0.0932,
+      "step": 9630
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 8.729640391331005e-06,
+      "loss": 0.1166,
+      "step": 9631
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 8.729507104203878e-06,
+      "loss": 0.1057,
+      "step": 9632
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 8.729373817076747e-06,
+      "loss": 0.0935,
+      "step": 9633
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 8.729240529949618e-06,
+      "loss": 0.1263,
+      "step": 9634
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 8.729107242822489e-06,
+      "loss": 0.0692,
+      "step": 9635
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 8.72897395569536e-06,
+      "loss": 0.0876,
+      "step": 9636
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 8.728840668568231e-06,
+      "loss": 0.1037,
+      "step": 9637
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 8.728707381441102e-06,
+      "loss": 0.0731,
+      "step": 9638
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 8.728574094313971e-06,
+      "loss": 0.0981,
+      "step": 9639
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 8.728440807186842e-06,
+      "loss": 0.1174,
+      "step": 9640
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 8.728307520059713e-06,
+      "loss": 0.1021,
+      "step": 9641
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 8.728174232932584e-06,
+      "loss": 0.0972,
+      "step": 9642
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 8.728040945805455e-06,
+      "loss": 0.1125,
+      "step": 9643
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 8.727907658678325e-06,
+      "loss": 0.1189,
+      "step": 9644
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 8.727774371551196e-06,
+      "loss": 0.1109,
+      "step": 9645
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 8.727641084424067e-06,
+      "loss": 0.0971,
+      "step": 9646
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 8.727507797296938e-06,
+      "loss": 0.1047,
+      "step": 9647
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 8.727374510169809e-06,
+      "loss": 0.1257,
+      "step": 9648
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 8.72724122304268e-06,
+      "loss": 0.0909,
+      "step": 9649
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 8.727107935915549e-06,
+      "loss": 0.0768,
+      "step": 9650
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 8.726974648788422e-06,
+      "loss": 0.1044,
+      "step": 9651
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 8.726841361661291e-06,
+      "loss": 0.1246,
+      "step": 9652
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 8.726708074534162e-06,
+      "loss": 0.1315,
+      "step": 9653
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 8.726574787407033e-06,
+      "loss": 0.0589,
+      "step": 9654
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 8.726441500279904e-06,
+      "loss": 0.1151,
+      "step": 9655
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 8.726308213152775e-06,
+      "loss": 0.1195,
+      "step": 9656
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 8.726174926025646e-06,
+      "loss": 0.1392,
+      "step": 9657
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 8.726041638898515e-06,
+      "loss": 0.1031,
+      "step": 9658
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 8.725908351771388e-06,
+      "loss": 0.146,
+      "step": 9659
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 8.725775064644257e-06,
+      "loss": 0.1186,
+      "step": 9660
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 8.725641777517128e-06,
+      "loss": 0.1584,
+      "step": 9661
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 8.72550849039e-06,
+      "loss": 0.0896,
+      "step": 9662
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 8.725375203262868e-06,
+      "loss": 0.0766,
+      "step": 9663
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 8.725241916135741e-06,
+      "loss": 0.1208,
+      "step": 9664
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 8.72510862900861e-06,
+      "loss": 0.1134,
+      "step": 9665
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 8.724975341881481e-06,
+      "loss": 0.0896,
+      "step": 9666
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 8.724842054754352e-06,
+      "loss": 0.1427,
+      "step": 9667
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 8.724708767627223e-06,
+      "loss": 0.0662,
+      "step": 9668
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 8.724575480500094e-06,
+      "loss": 0.0888,
+      "step": 9669
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 8.724442193372965e-06,
+      "loss": 0.1043,
+      "step": 9670
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 8.724308906245835e-06,
+      "loss": 0.0819,
+      "step": 9671
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 8.724175619118706e-06,
+      "loss": 0.1081,
+      "step": 9672
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 8.724042331991577e-06,
+      "loss": 0.0937,
+      "step": 9673
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 8.723909044864448e-06,
+      "loss": 0.0951,
+      "step": 9674
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 8.723775757737319e-06,
+      "loss": 0.1321,
+      "step": 9675
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 8.72364247061019e-06,
+      "loss": 0.0992,
+      "step": 9676
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 8.723509183483059e-06,
+      "loss": 0.0756,
+      "step": 9677
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 8.723375896355932e-06,
+      "loss": 0.0777,
+      "step": 9678
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 8.723242609228801e-06,
+      "loss": 0.1624,
+      "step": 9679
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 8.723109322101672e-06,
+      "loss": 0.142,
+      "step": 9680
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 8.722976034974543e-06,
+      "loss": 0.1132,
+      "step": 9681
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 8.722842747847412e-06,
+      "loss": 0.0838,
+      "step": 9682
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 8.722709460720285e-06,
+      "loss": 0.106,
+      "step": 9683
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 8.722576173593154e-06,
+      "loss": 0.0959,
+      "step": 9684
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 8.722442886466025e-06,
+      "loss": 0.1047,
+      "step": 9685
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 8.722309599338896e-06,
+      "loss": 0.1376,
+      "step": 9686
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 8.722176312211767e-06,
+      "loss": 0.0963,
+      "step": 9687
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 8.722043025084638e-06,
+      "loss": 0.1085,
+      "step": 9688
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 8.72190973795751e-06,
+      "loss": 0.1283,
+      "step": 9689
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 8.721776450830379e-06,
+      "loss": 0.1115,
+      "step": 9690
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 8.721643163703251e-06,
+      "loss": 0.1058,
+      "step": 9691
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 8.72150987657612e-06,
+      "loss": 0.1075,
+      "step": 9692
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 8.721376589448992e-06,
+      "loss": 0.1069,
+      "step": 9693
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 8.721243302321863e-06,
+      "loss": 0.1001,
+      "step": 9694
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 8.721110015194734e-06,
+      "loss": 0.1365,
+      "step": 9695
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 8.720976728067605e-06,
+      "loss": 0.0708,
+      "step": 9696
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 8.720843440940476e-06,
+      "loss": 0.0877,
+      "step": 9697
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 8.720710153813345e-06,
+      "loss": 0.0983,
+      "step": 9698
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 8.720576866686216e-06,
+      "loss": 0.0722,
+      "step": 9699
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 8.720443579559087e-06,
+      "loss": 0.1192,
+      "step": 9700
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 8.720310292431958e-06,
+      "loss": 0.0722,
+      "step": 9701
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 8.720177005304829e-06,
+      "loss": 0.1065,
+      "step": 9702
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 8.720043718177698e-06,
+      "loss": 0.1339,
+      "step": 9703
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 8.71991043105057e-06,
+      "loss": 0.0949,
+      "step": 9704
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 8.71977714392344e-06,
+      "loss": 0.1254,
+      "step": 9705
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 8.719643856796311e-06,
+      "loss": 0.1203,
+      "step": 9706
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 8.719510569669182e-06,
+      "loss": 0.0877,
+      "step": 9707
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 8.719377282542053e-06,
+      "loss": 0.109,
+      "step": 9708
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 8.719243995414924e-06,
+      "loss": 0.0806,
+      "step": 9709
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 8.719110708287795e-06,
+      "loss": 0.1373,
+      "step": 9710
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 8.718977421160664e-06,
+      "loss": 0.1583,
+      "step": 9711
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 8.718844134033535e-06,
+      "loss": 0.1371,
+      "step": 9712
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 8.718710846906406e-06,
+      "loss": 0.092,
+      "step": 9713
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 8.718577559779277e-06,
+      "loss": 0.1021,
+      "step": 9714
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 8.718444272652148e-06,
+      "loss": 0.1234,
+      "step": 9715
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 8.71831098552502e-06,
+      "loss": 0.0832,
+      "step": 9716
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 8.718177698397889e-06,
+      "loss": 0.0951,
+      "step": 9717
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 8.718044411270761e-06,
+      "loss": 0.0844,
+      "step": 9718
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 8.71791112414363e-06,
+      "loss": 0.1247,
+      "step": 9719
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 8.717777837016502e-06,
+      "loss": 0.0842,
+      "step": 9720
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 8.717644549889373e-06,
+      "loss": 0.1441,
+      "step": 9721
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 8.717511262762242e-06,
+      "loss": 0.1027,
+      "step": 9722
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 8.717377975635115e-06,
+      "loss": 0.1096,
+      "step": 9723
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 8.717244688507984e-06,
+      "loss": 0.0884,
+      "step": 9724
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 8.717111401380855e-06,
+      "loss": 0.1287,
+      "step": 9725
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 8.716978114253726e-06,
+      "loss": 0.1201,
+      "step": 9726
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 8.716844827126597e-06,
+      "loss": 0.0792,
+      "step": 9727
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 8.716711539999468e-06,
+      "loss": 0.1137,
+      "step": 9728
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 8.716578252872339e-06,
+      "loss": 0.109,
+      "step": 9729
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 8.716444965745208e-06,
+      "loss": 0.1197,
+      "step": 9730
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 8.716311678618081e-06,
+      "loss": 0.0913,
+      "step": 9731
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 8.71617839149095e-06,
+      "loss": 0.077,
+      "step": 9732
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 8.716045104363821e-06,
+      "loss": 0.1154,
+      "step": 9733
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 8.715911817236692e-06,
+      "loss": 0.1007,
+      "step": 9734
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 8.715778530109563e-06,
+      "loss": 0.0862,
+      "step": 9735
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 8.715645242982434e-06,
+      "loss": 0.0847,
+      "step": 9736
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 8.715511955855305e-06,
+      "loss": 0.0777,
+      "step": 9737
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 8.715378668728174e-06,
+      "loss": 0.1442,
+      "step": 9738
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 8.715245381601045e-06,
+      "loss": 0.0669,
+      "step": 9739
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 8.715112094473916e-06,
+      "loss": 0.0753,
+      "step": 9740
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 8.714978807346787e-06,
+      "loss": 0.0804,
+      "step": 9741
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 8.714845520219658e-06,
+      "loss": 0.127,
+      "step": 9742
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 8.714712233092528e-06,
+      "loss": 0.1399,
+      "step": 9743
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 8.714578945965399e-06,
+      "loss": 0.1282,
+      "step": 9744
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 8.71444565883827e-06,
+      "loss": 0.1082,
+      "step": 9745
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 8.71431237171114e-06,
+      "loss": 0.1197,
+      "step": 9746
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 8.714179084584012e-06,
+      "loss": 0.0946,
+      "step": 9747
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 8.714045797456883e-06,
+      "loss": 0.1082,
+      "step": 9748
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 8.713912510329752e-06,
+      "loss": 0.0976,
+      "step": 9749
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 8.713779223202625e-06,
+      "loss": 0.125,
+      "step": 9750
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 8.713645936075494e-06,
+      "loss": 0.0881,
+      "step": 9751
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 8.713512648948365e-06,
+      "loss": 0.0797,
+      "step": 9752
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 8.713379361821236e-06,
+      "loss": 0.0905,
+      "step": 9753
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 8.713246074694107e-06,
+      "loss": 0.079,
+      "step": 9754
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 8.713112787566978e-06,
+      "loss": 0.1172,
+      "step": 9755
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 8.712979500439849e-06,
+      "loss": 0.1361,
+      "step": 9756
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 8.712846213312718e-06,
+      "loss": 0.1089,
+      "step": 9757
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 8.71271292618559e-06,
+      "loss": 0.1248,
+      "step": 9758
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 8.71257963905846e-06,
+      "loss": 0.0951,
+      "step": 9759
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 8.712446351931331e-06,
+      "loss": 0.1001,
+      "step": 9760
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 8.712313064804202e-06,
+      "loss": 0.1516,
+      "step": 9761
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 8.712179777677072e-06,
+      "loss": 0.0809,
+      "step": 9762
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 8.712046490549944e-06,
+      "loss": 0.1037,
+      "step": 9763
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 8.711913203422814e-06,
+      "loss": 0.1241,
+      "step": 9764
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 8.711779916295685e-06,
+      "loss": 0.0865,
+      "step": 9765
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 8.711646629168556e-06,
+      "loss": 0.0988,
+      "step": 9766
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 8.711513342041427e-06,
+      "loss": 0.108,
+      "step": 9767
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 8.711380054914298e-06,
+      "loss": 0.117,
+      "step": 9768
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 8.711246767787169e-06,
+      "loss": 0.0912,
+      "step": 9769
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 8.711113480660038e-06,
+      "loss": 0.0811,
+      "step": 9770
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 8.71098019353291e-06,
+      "loss": 0.0875,
+      "step": 9771
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 8.71084690640578e-06,
+      "loss": 0.1291,
+      "step": 9772
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 8.71071361927865e-06,
+      "loss": 0.0805,
+      "step": 9773
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 8.710580332151522e-06,
+      "loss": 0.1102,
+      "step": 9774
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 8.710447045024393e-06,
+      "loss": 0.0843,
+      "step": 9775
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 8.710313757897264e-06,
+      "loss": 0.1487,
+      "step": 9776
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 8.710180470770133e-06,
+      "loss": 0.1285,
+      "step": 9777
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 8.710047183643004e-06,
+      "loss": 0.1017,
+      "step": 9778
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 8.709913896515875e-06,
+      "loss": 0.1216,
+      "step": 9779
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 8.709780609388746e-06,
+      "loss": 0.136,
+      "step": 9780
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 8.709647322261617e-06,
+      "loss": 0.1052,
+      "step": 9781
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 8.709514035134488e-06,
+      "loss": 0.1007,
+      "step": 9782
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 8.709380748007357e-06,
+      "loss": 0.1159,
+      "step": 9783
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 8.709247460880228e-06,
+      "loss": 0.1355,
+      "step": 9784
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 8.7091141737531e-06,
+      "loss": 0.1202,
+      "step": 9785
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 8.70898088662597e-06,
+      "loss": 0.1039,
+      "step": 9786
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 8.708847599498841e-06,
+      "loss": 0.0729,
+      "step": 9787
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 8.708714312371712e-06,
+      "loss": 0.0955,
+      "step": 9788
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 8.708581025244582e-06,
+      "loss": 0.1143,
+      "step": 9789
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 8.708447738117454e-06,
+      "loss": 0.1166,
+      "step": 9790
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 8.708314450990324e-06,
+      "loss": 0.084,
+      "step": 9791
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 8.708181163863195e-06,
+      "loss": 0.0722,
+      "step": 9792
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 8.708047876736066e-06,
+      "loss": 0.1294,
+      "step": 9793
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 8.707914589608937e-06,
+      "loss": 0.113,
+      "step": 9794
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 8.707781302481808e-06,
+      "loss": 0.1396,
+      "step": 9795
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 8.707648015354679e-06,
+      "loss": 0.0828,
+      "step": 9796
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 8.707514728227548e-06,
+      "loss": 0.1113,
+      "step": 9797
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 8.707381441100419e-06,
+      "loss": 0.0859,
+      "step": 9798
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 8.70724815397329e-06,
+      "loss": 0.0932,
+      "step": 9799
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 8.707114866846161e-06,
+      "loss": 0.1028,
+      "step": 9800
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 8.706981579719032e-06,
+      "loss": 0.0942,
+      "step": 9801
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 8.706848292591901e-06,
+      "loss": 0.1268,
+      "step": 9802
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 8.706715005464774e-06,
+      "loss": 0.1095,
+      "step": 9803
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 8.706581718337643e-06,
+      "loss": 0.0779,
+      "step": 9804
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 8.706448431210514e-06,
+      "loss": 0.0881,
+      "step": 9805
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 8.706315144083385e-06,
+      "loss": 0.089,
+      "step": 9806
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 8.706181856956256e-06,
+      "loss": 0.1425,
+      "step": 9807
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 8.706048569829127e-06,
+      "loss": 0.1248,
+      "step": 9808
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 8.705915282701998e-06,
+      "loss": 0.1121,
+      "step": 9809
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 8.705781995574867e-06,
+      "loss": 0.086,
+      "step": 9810
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 8.705648708447738e-06,
+      "loss": 0.097,
+      "step": 9811
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 8.70551542132061e-06,
+      "loss": 0.0937,
+      "step": 9812
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 8.70538213419348e-06,
+      "loss": 0.1028,
+      "step": 9813
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 8.705248847066351e-06,
+      "loss": 0.1161,
+      "step": 9814
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 8.705115559939222e-06,
+      "loss": 0.1015,
+      "step": 9815
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 8.704982272812092e-06,
+      "loss": 0.0739,
+      "step": 9816
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 8.704848985684963e-06,
+      "loss": 0.1084,
+      "step": 9817
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 8.704715698557834e-06,
+      "loss": 0.0587,
+      "step": 9818
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 8.704582411430705e-06,
+      "loss": 0.1254,
+      "step": 9819
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 8.704449124303576e-06,
+      "loss": 0.124,
+      "step": 9820
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 8.704315837176445e-06,
+      "loss": 0.1007,
+      "step": 9821
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 8.704182550049318e-06,
+      "loss": 0.0977,
+      "step": 9822
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 8.704049262922187e-06,
+      "loss": 0.1323,
+      "step": 9823
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 8.703915975795058e-06,
+      "loss": 0.0797,
+      "step": 9824
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 8.703782688667929e-06,
+      "loss": 0.104,
+      "step": 9825
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 8.7036494015408e-06,
+      "loss": 0.1396,
+      "step": 9826
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 8.703516114413671e-06,
+      "loss": 0.1222,
+      "step": 9827
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 8.703382827286542e-06,
+      "loss": 0.1402,
+      "step": 9828
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 8.703249540159411e-06,
+      "loss": 0.1215,
+      "step": 9829
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 8.703116253032284e-06,
+      "loss": 0.1153,
+      "step": 9830
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 8.702982965905153e-06,
+      "loss": 0.1261,
+      "step": 9831
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 8.702849678778024e-06,
+      "loss": 0.0926,
+      "step": 9832
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 8.702716391650895e-06,
+      "loss": 0.0948,
+      "step": 9833
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 8.702583104523766e-06,
+      "loss": 0.0823,
+      "step": 9834
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 8.702449817396637e-06,
+      "loss": 0.1294,
+      "step": 9835
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 8.702316530269507e-06,
+      "loss": 0.1288,
+      "step": 9836
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 8.702183243142378e-06,
+      "loss": 0.1554,
+      "step": 9837
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 8.702049956015249e-06,
+      "loss": 0.1129,
+      "step": 9838
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 8.70191666888812e-06,
+      "loss": 0.1179,
+      "step": 9839
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 8.70178338176099e-06,
+      "loss": 0.1289,
+      "step": 9840
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 8.701650094633862e-06,
+      "loss": 0.0996,
+      "step": 9841
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 8.70151680750673e-06,
+      "loss": 0.1294,
+      "step": 9842
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 8.701383520379603e-06,
+      "loss": 0.1101,
+      "step": 9843
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 8.701250233252473e-06,
+      "loss": 0.0779,
+      "step": 9844
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 8.701116946125344e-06,
+      "loss": 0.1136,
+      "step": 9845
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 8.700983658998215e-06,
+      "loss": 0.0813,
+      "step": 9846
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 8.700850371871086e-06,
+      "loss": 0.086,
+      "step": 9847
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 8.700717084743957e-06,
+      "loss": 0.104,
+      "step": 9848
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 8.700583797616828e-06,
+      "loss": 0.1366,
+      "step": 9849
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 8.700450510489697e-06,
+      "loss": 0.0779,
+      "step": 9850
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 8.700317223362568e-06,
+      "loss": 0.1557,
+      "step": 9851
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 8.700183936235439e-06,
+      "loss": 0.0892,
+      "step": 9852
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 8.70005064910831e-06,
+      "loss": 0.1396,
+      "step": 9853
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 8.699917361981181e-06,
+      "loss": 0.0707,
+      "step": 9854
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 8.699784074854052e-06,
+      "loss": 0.1028,
+      "step": 9855
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 8.699650787726921e-06,
+      "loss": 0.1213,
+      "step": 9856
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 8.699517500599792e-06,
+      "loss": 0.0669,
+      "step": 9857
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 8.699384213472663e-06,
+      "loss": 0.0647,
+      "step": 9858
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 8.699250926345534e-06,
+      "loss": 0.0771,
+      "step": 9859
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 8.699117639218405e-06,
+      "loss": 0.1418,
+      "step": 9860
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 8.698984352091275e-06,
+      "loss": 0.1163,
+      "step": 9861
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 8.698851064964147e-06,
+      "loss": 0.1126,
+      "step": 9862
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 8.698717777837017e-06,
+      "loss": 0.0907,
+      "step": 9863
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 8.698584490709888e-06,
+      "loss": 0.0945,
+      "step": 9864
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 8.698451203582759e-06,
+      "loss": 0.1084,
+      "step": 9865
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 8.69831791645563e-06,
+      "loss": 0.1555,
+      "step": 9866
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 8.6981846293285e-06,
+      "loss": 0.0936,
+      "step": 9867
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 8.698051342201372e-06,
+      "loss": 0.149,
+      "step": 9868
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 8.697918055074241e-06,
+      "loss": 0.0884,
+      "step": 9869
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 8.697784767947114e-06,
+      "loss": 0.116,
+      "step": 9870
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 8.697651480819983e-06,
+      "loss": 0.0785,
+      "step": 9871
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 8.697518193692854e-06,
+      "loss": 0.1349,
+      "step": 9872
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 8.697384906565725e-06,
+      "loss": 0.0886,
+      "step": 9873
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 8.697251619438596e-06,
+      "loss": 0.1265,
+      "step": 9874
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 8.697118332311467e-06,
+      "loss": 0.1271,
+      "step": 9875
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 8.696985045184336e-06,
+      "loss": 0.1078,
+      "step": 9876
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 8.696851758057207e-06,
+      "loss": 0.1239,
+      "step": 9877
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 8.696718470930078e-06,
+      "loss": 0.1,
+      "step": 9878
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 8.696585183802949e-06,
+      "loss": 0.101,
+      "step": 9879
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 8.69645189667582e-06,
+      "loss": 0.0843,
+      "step": 9880
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 8.696318609548691e-06,
+      "loss": 0.1131,
+      "step": 9881
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 8.69618532242156e-06,
+      "loss": 0.1249,
+      "step": 9882
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 8.696052035294431e-06,
+      "loss": 0.1354,
+      "step": 9883
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 8.695918748167302e-06,
+      "loss": 0.11,
+      "step": 9884
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 8.695785461040173e-06,
+      "loss": 0.0967,
+      "step": 9885
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 8.695652173913044e-06,
+      "loss": 0.0927,
+      "step": 9886
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 8.695518886785915e-06,
+      "loss": 0.0743,
+      "step": 9887
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 8.695385599658785e-06,
+      "loss": 0.1111,
+      "step": 9888
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 8.695252312531657e-06,
+      "loss": 0.1278,
+      "step": 9889
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 8.695119025404527e-06,
+      "loss": 0.0783,
+      "step": 9890
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 8.694985738277398e-06,
+      "loss": 0.088,
+      "step": 9891
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 8.694852451150269e-06,
+      "loss": 0.0803,
+      "step": 9892
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 8.69471916402314e-06,
+      "loss": 0.0988,
+      "step": 9893
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 8.69458587689601e-06,
+      "loss": 0.1289,
+      "step": 9894
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 8.69445258976888e-06,
+      "loss": 0.0752,
+      "step": 9895
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 8.694319302641751e-06,
+      "loss": 0.123,
+      "step": 9896
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 8.694186015514622e-06,
+      "loss": 0.1285,
+      "step": 9897
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 8.694052728387493e-06,
+      "loss": 0.0875,
+      "step": 9898
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 8.693919441260364e-06,
+      "loss": 0.1222,
+      "step": 9899
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 8.693786154133235e-06,
+      "loss": 0.104,
+      "step": 9900
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 8.693652867006104e-06,
+      "loss": 0.1122,
+      "step": 9901
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 8.693519579878977e-06,
+      "loss": 0.1142,
+      "step": 9902
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 8.693386292751846e-06,
+      "loss": 0.1696,
+      "step": 9903
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 8.693253005624717e-06,
+      "loss": 0.0759,
+      "step": 9904
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 8.693119718497588e-06,
+      "loss": 0.1187,
+      "step": 9905
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 8.69298643137046e-06,
+      "loss": 0.1191,
+      "step": 9906
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 8.69285314424333e-06,
+      "loss": 0.0779,
+      "step": 9907
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 8.692719857116201e-06,
+      "loss": 0.1143,
+      "step": 9908
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 8.69258656998907e-06,
+      "loss": 0.1007,
+      "step": 9909
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 8.692453282861942e-06,
+      "loss": 0.1136,
+      "step": 9910
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 8.692319995734813e-06,
+      "loss": 0.0862,
+      "step": 9911
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 8.692186708607683e-06,
+      "loss": 0.0942,
+      "step": 9912
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 8.692053421480554e-06,
+      "loss": 0.0984,
+      "step": 9913
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 8.691920134353425e-06,
+      "loss": 0.1231,
+      "step": 9914
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 8.691786847226295e-06,
+      "loss": 0.1135,
+      "step": 9915
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 8.691653560099166e-06,
+      "loss": 0.1211,
+      "step": 9916
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 8.691520272972037e-06,
+      "loss": 0.1134,
+      "step": 9917
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 8.691386985844908e-06,
+      "loss": 0.0924,
+      "step": 9918
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 8.691253698717779e-06,
+      "loss": 0.0842,
+      "step": 9919
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 8.691120411590648e-06,
+      "loss": 0.1034,
+      "step": 9920
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 8.69098712446352e-06,
+      "loss": 0.1101,
+      "step": 9921
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 8.69085383733639e-06,
+      "loss": 0.0869,
+      "step": 9922
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 8.690720550209261e-06,
+      "loss": 0.0702,
+      "step": 9923
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 8.690587263082132e-06,
+      "loss": 0.1024,
+      "step": 9924
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 8.690453975955003e-06,
+      "loss": 0.1246,
+      "step": 9925
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 8.690320688827874e-06,
+      "loss": 0.102,
+      "step": 9926
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 8.690187401700745e-06,
+      "loss": 0.0947,
+      "step": 9927
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 8.690054114573614e-06,
+      "loss": 0.1077,
+      "step": 9928
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 8.689920827446487e-06,
+      "loss": 0.0938,
+      "step": 9929
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 8.689787540319356e-06,
+      "loss": 0.0734,
+      "step": 9930
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 8.689654253192227e-06,
+      "loss": 0.0821,
+      "step": 9931
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 8.689520966065098e-06,
+      "loss": 0.0951,
+      "step": 9932
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 8.68938767893797e-06,
+      "loss": 0.0994,
+      "step": 9933
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 8.68925439181084e-06,
+      "loss": 0.1028,
+      "step": 9934
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 8.68912110468371e-06,
+      "loss": 0.1429,
+      "step": 9935
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 8.68898781755658e-06,
+      "loss": 0.1118,
+      "step": 9936
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 8.688854530429452e-06,
+      "loss": 0.0914,
+      "step": 9937
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 8.688721243302323e-06,
+      "loss": 0.1454,
+      "step": 9938
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 8.688587956175194e-06,
+      "loss": 0.0974,
+      "step": 9939
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 8.688454669048065e-06,
+      "loss": 0.0715,
+      "step": 9940
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 8.688321381920934e-06,
+      "loss": 0.1012,
+      "step": 9941
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 8.688188094793807e-06,
+      "loss": 0.1271,
+      "step": 9942
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 8.688054807666676e-06,
+      "loss": 0.1132,
+      "step": 9943
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 8.687921520539547e-06,
+      "loss": 0.1082,
+      "step": 9944
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 8.687788233412418e-06,
+      "loss": 0.1248,
+      "step": 9945
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 8.687654946285289e-06,
+      "loss": 0.0747,
+      "step": 9946
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 8.68752165915816e-06,
+      "loss": 0.1162,
+      "step": 9947
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 8.68738837203103e-06,
+      "loss": 0.0883,
+      "step": 9948
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 8.6872550849039e-06,
+      "loss": 0.1559,
+      "step": 9949
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 8.687121797776771e-06,
+      "loss": 0.1184,
+      "step": 9950
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 8.686988510649642e-06,
+      "loss": 0.1192,
+      "step": 9951
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 8.686855223522513e-06,
+      "loss": 0.0739,
+      "step": 9952
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 8.686721936395384e-06,
+      "loss": 0.0676,
+      "step": 9953
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 8.686588649268253e-06,
+      "loss": 0.077,
+      "step": 9954
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 8.686455362141124e-06,
+      "loss": 0.1467,
+      "step": 9955
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 8.686322075013995e-06,
+      "loss": 0.1094,
+      "step": 9956
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 8.686188787886866e-06,
+      "loss": 0.0742,
+      "step": 9957
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 8.686055500759737e-06,
+      "loss": 0.124,
+      "step": 9958
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 8.685922213632608e-06,
+      "loss": 0.0783,
+      "step": 9959
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 8.685788926505478e-06,
+      "loss": 0.1133,
+      "step": 9960
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 8.68565563937835e-06,
+      "loss": 0.0974,
+      "step": 9961
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 8.68552235225122e-06,
+      "loss": 0.0872,
+      "step": 9962
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 8.68538906512409e-06,
+      "loss": 0.1152,
+      "step": 9963
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 8.685255777996962e-06,
+      "loss": 0.0997,
+      "step": 9964
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 8.685122490869833e-06,
+      "loss": 0.0993,
+      "step": 9965
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 8.684989203742704e-06,
+      "loss": 0.0799,
+      "step": 9966
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 8.684855916615575e-06,
+      "loss": 0.1311,
+      "step": 9967
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 8.684722629488444e-06,
+      "loss": 0.0811,
+      "step": 9968
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 8.684589342361317e-06,
+      "loss": 0.1305,
+      "step": 9969
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 8.684456055234186e-06,
+      "loss": 0.0866,
+      "step": 9970
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 8.684322768107057e-06,
+      "loss": 0.1016,
+      "step": 9971
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 8.684189480979928e-06,
+      "loss": 0.0838,
+      "step": 9972
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 8.684056193852797e-06,
+      "loss": 0.1438,
+      "step": 9973
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 8.68392290672567e-06,
+      "loss": 0.1415,
+      "step": 9974
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 8.68378961959854e-06,
+      "loss": 0.0841,
+      "step": 9975
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 8.68365633247141e-06,
+      "loss": 0.1203,
+      "step": 9976
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 8.683523045344281e-06,
+      "loss": 0.099,
+      "step": 9977
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 8.683389758217152e-06,
+      "loss": 0.1035,
+      "step": 9978
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 8.683256471090023e-06,
+      "loss": 0.0726,
+      "step": 9979
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 8.683123183962894e-06,
+      "loss": 0.0818,
+      "step": 9980
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 8.682989896835763e-06,
+      "loss": 0.1204,
+      "step": 9981
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 8.682856609708634e-06,
+      "loss": 0.1153,
+      "step": 9982
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 8.682723322581505e-06,
+      "loss": 0.1307,
+      "step": 9983
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 8.682590035454376e-06,
+      "loss": 0.0914,
+      "step": 9984
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 8.682456748327247e-06,
+      "loss": 0.0926,
+      "step": 9985
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 8.682323461200118e-06,
+      "loss": 0.084,
+      "step": 9986
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 8.682190174072988e-06,
+      "loss": 0.107,
+      "step": 9987
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 8.68205688694586e-06,
+      "loss": 0.0938,
+      "step": 9988
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 8.68192359981873e-06,
+      "loss": 0.1019,
+      "step": 9989
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 8.6817903126916e-06,
+      "loss": 0.0637,
+      "step": 9990
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 8.681657025564472e-06,
+      "loss": 0.0852,
+      "step": 9991
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 8.681523738437343e-06,
+      "loss": 0.093,
+      "step": 9992
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 8.681390451310214e-06,
+      "loss": 0.0926,
+      "step": 9993
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 8.681257164183083e-06,
+      "loss": 0.1244,
+      "step": 9994
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 8.681123877055954e-06,
+      "loss": 0.1055,
+      "step": 9995
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 8.680990589928825e-06,
+      "loss": 0.0902,
+      "step": 9996
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 8.680857302801696e-06,
+      "loss": 0.107,
+      "step": 9997
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 8.680724015674567e-06,
+      "loss": 0.1231,
+      "step": 9998
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 8.680590728547438e-06,
+      "loss": 0.0948,
+      "step": 9999
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 8.680457441420307e-06,
+      "loss": 0.1674,
+      "step": 10000
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 8.68032415429318e-06,
+      "loss": 0.097,
+      "step": 10001
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 8.68019086716605e-06,
+      "loss": 0.1157,
+      "step": 10002
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 8.68005758003892e-06,
+      "loss": 0.1091,
+      "step": 10003
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 8.679924292911791e-06,
+      "loss": 0.0884,
+      "step": 10004
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 8.679791005784662e-06,
+      "loss": 0.0843,
+      "step": 10005
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 8.679657718657533e-06,
+      "loss": 0.0919,
+      "step": 10006
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 8.679524431530404e-06,
+      "loss": 0.1293,
+      "step": 10007
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 8.679391144403274e-06,
+      "loss": 0.0848,
+      "step": 10008
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 8.679257857276146e-06,
+      "loss": 0.1387,
+      "step": 10009
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 8.679124570149016e-06,
+      "loss": 0.0874,
+      "step": 10010
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 8.678991283021887e-06,
+      "loss": 0.1224,
+      "step": 10011
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 8.678857995894758e-06,
+      "loss": 0.1638,
+      "step": 10012
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 8.678724708767627e-06,
+      "loss": 0.1059,
+      "step": 10013
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 8.6785914216405e-06,
+      "loss": 0.1249,
+      "step": 10014
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 8.678458134513369e-06,
+      "loss": 0.1155,
+      "step": 10015
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 8.67832484738624e-06,
+      "loss": 0.1041,
+      "step": 10016
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 8.67819156025911e-06,
+      "loss": 0.0958,
+      "step": 10017
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 8.678058273131982e-06,
+      "loss": 0.0801,
+      "step": 10018
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 8.677924986004853e-06,
+      "loss": 0.0901,
+      "step": 10019
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 8.677791698877724e-06,
+      "loss": 0.1323,
+      "step": 10020
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 8.677658411750593e-06,
+      "loss": 0.1527,
+      "step": 10021
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 8.677525124623464e-06,
+      "loss": 0.1201,
+      "step": 10022
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 8.677391837496335e-06,
+      "loss": 0.125,
+      "step": 10023
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 8.677258550369206e-06,
+      "loss": 0.1088,
+      "step": 10024
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 8.677125263242077e-06,
+      "loss": 0.1185,
+      "step": 10025
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 8.676991976114948e-06,
+      "loss": 0.0731,
+      "step": 10026
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 8.676858688987817e-06,
+      "loss": 0.0915,
+      "step": 10027
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 8.67672540186069e-06,
+      "loss": 0.1034,
+      "step": 10028
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 8.67659211473356e-06,
+      "loss": 0.0872,
+      "step": 10029
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 8.67645882760643e-06,
+      "loss": 0.0649,
+      "step": 10030
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 8.676325540479301e-06,
+      "loss": 0.0871,
+      "step": 10031
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 8.67619225335217e-06,
+      "loss": 0.1059,
+      "step": 10032
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 8.676058966225043e-06,
+      "loss": 0.1087,
+      "step": 10033
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 8.675925679097913e-06,
+      "loss": 0.0914,
+      "step": 10034
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 8.675792391970784e-06,
+      "loss": 0.1246,
+      "step": 10035
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 8.675659104843655e-06,
+      "loss": 0.1023,
+      "step": 10036
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 8.675525817716526e-06,
+      "loss": 0.1103,
+      "step": 10037
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 8.675392530589397e-06,
+      "loss": 0.0854,
+      "step": 10038
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 8.675259243462268e-06,
+      "loss": 0.0936,
+      "step": 10039
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 8.675125956335137e-06,
+      "loss": 0.0697,
+      "step": 10040
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 8.67499266920801e-06,
+      "loss": 0.0918,
+      "step": 10041
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 8.674859382080879e-06,
+      "loss": 0.0823,
+      "step": 10042
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 8.67472609495375e-06,
+      "loss": 0.1247,
+      "step": 10043
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 8.674592807826621e-06,
+      "loss": 0.092,
+      "step": 10044
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 8.674459520699492e-06,
+      "loss": 0.0987,
+      "step": 10045
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 8.674326233572363e-06,
+      "loss": 0.0706,
+      "step": 10046
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 8.674192946445234e-06,
+      "loss": 0.1729,
+      "step": 10047
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 8.674059659318103e-06,
+      "loss": 0.1165,
+      "step": 10048
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 8.673926372190974e-06,
+      "loss": 0.1191,
+      "step": 10049
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 8.673793085063845e-06,
+      "loss": 0.1195,
+      "step": 10050
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 8.673659797936716e-06,
+      "loss": 0.1022,
+      "step": 10051
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 8.673526510809587e-06,
+      "loss": 0.1318,
+      "step": 10052
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 8.673393223682456e-06,
+      "loss": 0.0893,
+      "step": 10053
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 8.673259936555327e-06,
+      "loss": 0.1466,
+      "step": 10054
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 8.673126649428198e-06,
+      "loss": 0.0855,
+      "step": 10055
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 8.67299336230107e-06,
+      "loss": 0.1236,
+      "step": 10056
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 8.67286007517394e-06,
+      "loss": 0.1024,
+      "step": 10057
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 8.672726788046811e-06,
+      "loss": 0.066,
+      "step": 10058
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 8.67259350091968e-06,
+      "loss": 0.1128,
+      "step": 10059
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 8.672460213792553e-06,
+      "loss": 0.0724,
+      "step": 10060
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 8.672326926665423e-06,
+      "loss": 0.1026,
+      "step": 10061
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 8.672193639538294e-06,
+      "loss": 0.0994,
+      "step": 10062
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 8.672060352411165e-06,
+      "loss": 0.1413,
+      "step": 10063
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 8.671927065284036e-06,
+      "loss": 0.1166,
+      "step": 10064
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 8.671793778156907e-06,
+      "loss": 0.0868,
+      "step": 10065
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 8.671660491029778e-06,
+      "loss": 0.0705,
+      "step": 10066
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 8.671527203902647e-06,
+      "loss": 0.0895,
+      "step": 10067
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 8.67139391677552e-06,
+      "loss": 0.1219,
+      "step": 10068
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 8.671260629648389e-06,
+      "loss": 0.109,
+      "step": 10069
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 8.67112734252126e-06,
+      "loss": 0.0925,
+      "step": 10070
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 8.670994055394131e-06,
+      "loss": 0.096,
+      "step": 10071
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 8.670860768267e-06,
+      "loss": 0.0876,
+      "step": 10072
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 8.670727481139873e-06,
+      "loss": 0.1347,
+      "step": 10073
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 8.670594194012742e-06,
+      "loss": 0.064,
+      "step": 10074
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 8.670460906885613e-06,
+      "loss": 0.1169,
+      "step": 10075
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 8.670327619758484e-06,
+      "loss": 0.1184,
+      "step": 10076
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 8.670194332631355e-06,
+      "loss": 0.1407,
+      "step": 10077
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 8.670061045504226e-06,
+      "loss": 0.1624,
+      "step": 10078
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 8.669927758377097e-06,
+      "loss": 0.0912,
+      "step": 10079
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 8.669794471249967e-06,
+      "loss": 0.1324,
+      "step": 10080
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 8.66966118412284e-06,
+      "loss": 0.1366,
+      "step": 10081
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 8.669527896995709e-06,
+      "loss": 0.1308,
+      "step": 10082
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 8.66939460986858e-06,
+      "loss": 0.1252,
+      "step": 10083
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 8.66926132274145e-06,
+      "loss": 0.1096,
+      "step": 10084
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 8.669128035614322e-06,
+      "loss": 0.1263,
+      "step": 10085
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 8.668994748487193e-06,
+      "loss": 0.1117,
+      "step": 10086
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 8.668861461360064e-06,
+      "loss": 0.1674,
+      "step": 10087
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 8.668728174232933e-06,
+      "loss": 0.0942,
+      "step": 10088
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 8.668594887105804e-06,
+      "loss": 0.0657,
+      "step": 10089
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 8.668461599978675e-06,
+      "loss": 0.1185,
+      "step": 10090
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 8.668328312851546e-06,
+      "loss": 0.0753,
+      "step": 10091
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 8.668195025724417e-06,
+      "loss": 0.1127,
+      "step": 10092
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 8.668061738597286e-06,
+      "loss": 0.1547,
+      "step": 10093
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 8.667928451470157e-06,
+      "loss": 0.1166,
+      "step": 10094
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 8.667795164343028e-06,
+      "loss": 0.1482,
+      "step": 10095
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 8.667661877215899e-06,
+      "loss": 0.1131,
+      "step": 10096
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 8.66752859008877e-06,
+      "loss": 0.097,
+      "step": 10097
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 8.667395302961641e-06,
+      "loss": 0.1146,
+      "step": 10098
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 8.66726201583451e-06,
+      "loss": 0.1092,
+      "step": 10099
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 8.667128728707383e-06,
+      "loss": 0.0724,
+      "step": 10100
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 8.666995441580252e-06,
+      "loss": 0.0904,
+      "step": 10101
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 8.666862154453123e-06,
+      "loss": 0.0934,
+      "step": 10102
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 8.666728867325994e-06,
+      "loss": 0.1214,
+      "step": 10103
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 8.666595580198865e-06,
+      "loss": 0.1236,
+      "step": 10104
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 8.666462293071736e-06,
+      "loss": 0.0855,
+      "step": 10105
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 8.666329005944607e-06,
+      "loss": 0.1585,
+      "step": 10106
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 8.666195718817477e-06,
+      "loss": 0.1092,
+      "step": 10107
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 8.66606243169035e-06,
+      "loss": 0.0938,
+      "step": 10108
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 8.665929144563219e-06,
+      "loss": 0.0637,
+      "step": 10109
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 8.66579585743609e-06,
+      "loss": 0.1196,
+      "step": 10110
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 8.66566257030896e-06,
+      "loss": 0.0816,
+      "step": 10111
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 8.66552928318183e-06,
+      "loss": 0.1137,
+      "step": 10112
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 8.665395996054703e-06,
+      "loss": 0.1129,
+      "step": 10113
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 8.665262708927572e-06,
+      "loss": 0.1175,
+      "step": 10114
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 8.665129421800443e-06,
+      "loss": 0.0905,
+      "step": 10115
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 8.664996134673314e-06,
+      "loss": 0.0825,
+      "step": 10116
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 8.664862847546185e-06,
+      "loss": 0.0884,
+      "step": 10117
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 8.664729560419056e-06,
+      "loss": 0.0934,
+      "step": 10118
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 8.664596273291927e-06,
+      "loss": 0.068,
+      "step": 10119
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 8.664462986164796e-06,
+      "loss": 0.13,
+      "step": 10120
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 8.664329699037667e-06,
+      "loss": 0.1051,
+      "step": 10121
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 8.664196411910538e-06,
+      "loss": 0.1089,
+      "step": 10122
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 8.66406312478341e-06,
+      "loss": 0.0839,
+      "step": 10123
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 8.66392983765628e-06,
+      "loss": 0.112,
+      "step": 10124
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 8.663796550529151e-06,
+      "loss": 0.0908,
+      "step": 10125
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 8.66366326340202e-06,
+      "loss": 0.1146,
+      "step": 10126
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 8.663529976274893e-06,
+      "loss": 0.1004,
+      "step": 10127
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 8.663396689147762e-06,
+      "loss": 0.0707,
+      "step": 10128
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 8.663263402020633e-06,
+      "loss": 0.1433,
+      "step": 10129
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 8.663130114893504e-06,
+      "loss": 0.0908,
+      "step": 10130
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 8.662996827766374e-06,
+      "loss": 0.0834,
+      "step": 10131
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 8.662863540639246e-06,
+      "loss": 0.0996,
+      "step": 10132
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 8.662730253512116e-06,
+      "loss": 0.1129,
+      "step": 10133
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 8.662596966384987e-06,
+      "loss": 0.1219,
+      "step": 10134
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 8.662463679257858e-06,
+      "loss": 0.128,
+      "step": 10135
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 8.662330392130729e-06,
+      "loss": 0.0776,
+      "step": 10136
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 8.6621971050036e-06,
+      "loss": 0.0804,
+      "step": 10137
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 8.66206381787647e-06,
+      "loss": 0.0975,
+      "step": 10138
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 8.66193053074934e-06,
+      "loss": 0.1232,
+      "step": 10139
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 8.661797243622213e-06,
+      "loss": 0.1212,
+      "step": 10140
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 8.661663956495082e-06,
+      "loss": 0.0954,
+      "step": 10141
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 8.661530669367953e-06,
+      "loss": 0.0918,
+      "step": 10142
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 8.661397382240824e-06,
+      "loss": 0.0961,
+      "step": 10143
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 8.661264095113695e-06,
+      "loss": 0.1174,
+      "step": 10144
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 8.661130807986566e-06,
+      "loss": 0.1054,
+      "step": 10145
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 8.660997520859437e-06,
+      "loss": 0.1063,
+      "step": 10146
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 8.660864233732306e-06,
+      "loss": 0.1078,
+      "step": 10147
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 8.660730946605177e-06,
+      "loss": 0.1286,
+      "step": 10148
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 8.660597659478048e-06,
+      "loss": 0.096,
+      "step": 10149
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 8.66046437235092e-06,
+      "loss": 0.1009,
+      "step": 10150
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 8.66033108522379e-06,
+      "loss": 0.0965,
+      "step": 10151
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 8.66019779809666e-06,
+      "loss": 0.0964,
+      "step": 10152
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 8.66006451096953e-06,
+      "loss": 0.1321,
+      "step": 10153
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 8.659931223842402e-06,
+      "loss": 0.122,
+      "step": 10154
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 8.659797936715273e-06,
+      "loss": 0.101,
+      "step": 10155
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 8.659664649588144e-06,
+      "loss": 0.0852,
+      "step": 10156
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 8.659531362461015e-06,
+      "loss": 0.1237,
+      "step": 10157
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 8.659398075333884e-06,
+      "loss": 0.092,
+      "step": 10158
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 8.659264788206757e-06,
+      "loss": 0.1017,
+      "step": 10159
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 8.659131501079626e-06,
+      "loss": 0.1255,
+      "step": 10160
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 8.658998213952497e-06,
+      "loss": 0.1067,
+      "step": 10161
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 8.658864926825368e-06,
+      "loss": 0.1575,
+      "step": 10162
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 8.658731639698239e-06,
+      "loss": 0.108,
+      "step": 10163
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 8.65859835257111e-06,
+      "loss": 0.1044,
+      "step": 10164
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 8.65846506544398e-06,
+      "loss": 0.1561,
+      "step": 10165
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 8.65833177831685e-06,
+      "loss": 0.1236,
+      "step": 10166
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 8.658198491189723e-06,
+      "loss": 0.1452,
+      "step": 10167
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 8.658065204062592e-06,
+      "loss": 0.11,
+      "step": 10168
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 8.657931916935463e-06,
+      "loss": 0.0739,
+      "step": 10169
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 8.657798629808334e-06,
+      "loss": 0.1583,
+      "step": 10170
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 8.657665342681203e-06,
+      "loss": 0.0795,
+      "step": 10171
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 8.657532055554076e-06,
+      "loss": 0.1179,
+      "step": 10172
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 8.657398768426945e-06,
+      "loss": 0.0934,
+      "step": 10173
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 8.657265481299816e-06,
+      "loss": 0.1027,
+      "step": 10174
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 8.657132194172687e-06,
+      "loss": 0.092,
+      "step": 10175
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 8.656998907045558e-06,
+      "loss": 0.1312,
+      "step": 10176
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 8.65686561991843e-06,
+      "loss": 0.0708,
+      "step": 10177
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 8.6567323327913e-06,
+      "loss": 0.1226,
+      "step": 10178
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 8.65659904566417e-06,
+      "loss": 0.0859,
+      "step": 10179
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 8.656465758537042e-06,
+      "loss": 0.0967,
+      "step": 10180
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 8.656332471409912e-06,
+      "loss": 0.0751,
+      "step": 10181
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 8.656199184282783e-06,
+      "loss": 0.0843,
+      "step": 10182
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 8.656065897155654e-06,
+      "loss": 0.118,
+      "step": 10183
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 8.655932610028525e-06,
+      "loss": 0.089,
+      "step": 10184
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 8.655799322901396e-06,
+      "loss": 0.0865,
+      "step": 10185
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 8.655666035774267e-06,
+      "loss": 0.0834,
+      "step": 10186
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 8.655532748647136e-06,
+      "loss": 0.1241,
+      "step": 10187
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 8.655399461520007e-06,
+      "loss": 0.1102,
+      "step": 10188
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 8.655266174392878e-06,
+      "loss": 0.0923,
+      "step": 10189
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 8.655132887265749e-06,
+      "loss": 0.1185,
+      "step": 10190
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 8.65499960013862e-06,
+      "loss": 0.1451,
+      "step": 10191
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 8.65486631301149e-06,
+      "loss": 0.0953,
+      "step": 10192
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 8.65473302588436e-06,
+      "loss": 0.0964,
+      "step": 10193
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 8.654599738757231e-06,
+      "loss": 0.1168,
+      "step": 10194
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 8.654466451630102e-06,
+      "loss": 0.0967,
+      "step": 10195
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 8.654333164502973e-06,
+      "loss": 0.0894,
+      "step": 10196
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 8.654199877375844e-06,
+      "loss": 0.0946,
+      "step": 10197
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 8.654066590248713e-06,
+      "loss": 0.1446,
+      "step": 10198
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 8.653933303121586e-06,
+      "loss": 0.0824,
+      "step": 10199
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 8.653800015994455e-06,
+      "loss": 0.1297,
+      "step": 10200
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 8.653666728867326e-06,
+      "loss": 0.0873,
+      "step": 10201
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 8.653533441740197e-06,
+      "loss": 0.0911,
+      "step": 10202
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 8.653400154613068e-06,
+      "loss": 0.1556,
+      "step": 10203
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 8.65326686748594e-06,
+      "loss": 0.0877,
+      "step": 10204
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 8.65313358035881e-06,
+      "loss": 0.1496,
+      "step": 10205
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 8.65300029323168e-06,
+      "loss": 0.115,
+      "step": 10206
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 8.652867006104552e-06,
+      "loss": 0.1468,
+      "step": 10207
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 8.652733718977422e-06,
+      "loss": 0.1072,
+      "step": 10208
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 8.652600431850293e-06,
+      "loss": 0.0829,
+      "step": 10209
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 8.652467144723164e-06,
+      "loss": 0.1495,
+      "step": 10210
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 8.652333857596033e-06,
+      "loss": 0.0844,
+      "step": 10211
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 8.652200570468906e-06,
+      "loss": 0.1096,
+      "step": 10212
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 8.652067283341775e-06,
+      "loss": 0.1236,
+      "step": 10213
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 8.651933996214646e-06,
+      "loss": 0.1037,
+      "step": 10214
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 8.651800709087517e-06,
+      "loss": 0.1202,
+      "step": 10215
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 8.651667421960388e-06,
+      "loss": 0.0974,
+      "step": 10216
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 8.651534134833259e-06,
+      "loss": 0.1074,
+      "step": 10217
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 8.65140084770613e-06,
+      "loss": 0.1136,
+      "step": 10218
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 8.651267560579e-06,
+      "loss": 0.1296,
+      "step": 10219
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 8.65113427345187e-06,
+      "loss": 0.1055,
+      "step": 10220
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 8.651000986324741e-06,
+      "loss": 0.0987,
+      "step": 10221
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 8.650867699197612e-06,
+      "loss": 0.0963,
+      "step": 10222
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 8.650734412070483e-06,
+      "loss": 0.1418,
+      "step": 10223
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 8.650601124943354e-06,
+      "loss": 0.1006,
+      "step": 10224
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 8.650467837816224e-06,
+      "loss": 0.1057,
+      "step": 10225
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 8.650334550689096e-06,
+      "loss": 0.1305,
+      "step": 10226
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 8.650201263561966e-06,
+      "loss": 0.084,
+      "step": 10227
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 8.650067976434837e-06,
+      "loss": 0.1244,
+      "step": 10228
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 8.649934689307708e-06,
+      "loss": 0.1083,
+      "step": 10229
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 8.649801402180577e-06,
+      "loss": 0.0992,
+      "step": 10230
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 8.64966811505345e-06,
+      "loss": 0.0837,
+      "step": 10231
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 8.649534827926319e-06,
+      "loss": 0.0816,
+      "step": 10232
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 8.64940154079919e-06,
+      "loss": 0.1393,
+      "step": 10233
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 8.64926825367206e-06,
+      "loss": 0.0829,
+      "step": 10234
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 8.649134966544932e-06,
+      "loss": 0.0668,
+      "step": 10235
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 8.649001679417803e-06,
+      "loss": 0.12,
+      "step": 10236
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 8.648868392290674e-06,
+      "loss": 0.0841,
+      "step": 10237
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 8.648735105163543e-06,
+      "loss": 0.0956,
+      "step": 10238
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 8.648601818036416e-06,
+      "loss": 0.1289,
+      "step": 10239
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 8.648468530909285e-06,
+      "loss": 0.0942,
+      "step": 10240
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 8.648335243782156e-06,
+      "loss": 0.1139,
+      "step": 10241
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 8.648201956655027e-06,
+      "loss": 0.1214,
+      "step": 10242
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 8.648068669527898e-06,
+      "loss": 0.1206,
+      "step": 10243
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 8.647935382400769e-06,
+      "loss": 0.0939,
+      "step": 10244
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 8.64780209527364e-06,
+      "loss": 0.1021,
+      "step": 10245
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 8.64766880814651e-06,
+      "loss": 0.1468,
+      "step": 10246
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 8.647535521019382e-06,
+      "loss": 0.1037,
+      "step": 10247
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 8.647402233892251e-06,
+      "loss": 0.101,
+      "step": 10248
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 8.647268946765122e-06,
+      "loss": 0.1021,
+      "step": 10249
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 8.647135659637993e-06,
+      "loss": 0.0807,
+      "step": 10250
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 8.647002372510863e-06,
+      "loss": 0.1109,
+      "step": 10251
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 8.646869085383735e-06,
+      "loss": 0.0983,
+      "step": 10252
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 8.646735798256605e-06,
+      "loss": 0.1068,
+      "step": 10253
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 8.646602511129476e-06,
+      "loss": 0.1016,
+      "step": 10254
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 8.646469224002347e-06,
+      "loss": 0.1086,
+      "step": 10255
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 8.646335936875218e-06,
+      "loss": 0.1236,
+      "step": 10256
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 8.646202649748089e-06,
+      "loss": 0.1028,
+      "step": 10257
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 8.64606936262096e-06,
+      "loss": 0.095,
+      "step": 10258
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 8.645936075493829e-06,
+      "loss": 0.1078,
+      "step": 10259
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 8.6458027883667e-06,
+      "loss": 0.1002,
+      "step": 10260
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 8.645669501239571e-06,
+      "loss": 0.1179,
+      "step": 10261
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 8.645536214112442e-06,
+      "loss": 0.1312,
+      "step": 10262
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 8.645402926985313e-06,
+      "loss": 0.0801,
+      "step": 10263
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 8.645269639858184e-06,
+      "loss": 0.1203,
+      "step": 10264
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 8.645136352731053e-06,
+      "loss": 0.1167,
+      "step": 10265
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 8.645003065603926e-06,
+      "loss": 0.0871,
+      "step": 10266
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 8.644869778476795e-06,
+      "loss": 0.1642,
+      "step": 10267
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 8.644736491349666e-06,
+      "loss": 0.0883,
+      "step": 10268
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 8.644603204222537e-06,
+      "loss": 0.1352,
+      "step": 10269
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 8.644469917095406e-06,
+      "loss": 0.0498,
+      "step": 10270
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 8.644336629968279e-06,
+      "loss": 0.1318,
+      "step": 10271
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 8.644203342841148e-06,
+      "loss": 0.1082,
+      "step": 10272
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 8.64407005571402e-06,
+      "loss": 0.1291,
+      "step": 10273
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 8.64393676858689e-06,
+      "loss": 0.1389,
+      "step": 10274
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 8.643803481459761e-06,
+      "loss": 0.1104,
+      "step": 10275
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 8.643670194332632e-06,
+      "loss": 0.1272,
+      "step": 10276
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 8.643536907205503e-06,
+      "loss": 0.0871,
+      "step": 10277
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 8.643403620078373e-06,
+      "loss": 0.0787,
+      "step": 10278
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 8.643270332951245e-06,
+      "loss": 0.1003,
+      "step": 10279
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 8.643137045824115e-06,
+      "loss": 0.0806,
+      "step": 10280
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 8.643003758696986e-06,
+      "loss": 0.1049,
+      "step": 10281
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 8.642870471569857e-06,
+      "loss": 0.1141,
+      "step": 10282
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 8.642737184442728e-06,
+      "loss": 0.1145,
+      "step": 10283
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 8.642603897315599e-06,
+      "loss": 0.1061,
+      "step": 10284
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 8.64247061018847e-06,
+      "loss": 0.0901,
+      "step": 10285
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 8.642337323061339e-06,
+      "loss": 0.1467,
+      "step": 10286
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 8.64220403593421e-06,
+      "loss": 0.133,
+      "step": 10287
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 8.642070748807081e-06,
+      "loss": 0.0924,
+      "step": 10288
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 8.641937461679952e-06,
+      "loss": 0.0919,
+      "step": 10289
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 8.641804174552823e-06,
+      "loss": 0.0829,
+      "step": 10290
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 8.641670887425692e-06,
+      "loss": 0.1022,
+      "step": 10291
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 8.641537600298563e-06,
+      "loss": 0.1007,
+      "step": 10292
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 8.641404313171434e-06,
+      "loss": 0.1112,
+      "step": 10293
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 8.641271026044305e-06,
+      "loss": 0.1237,
+      "step": 10294
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 8.641137738917176e-06,
+      "loss": 0.1452,
+      "step": 10295
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 8.641004451790047e-06,
+      "loss": 0.1498,
+      "step": 10296
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 8.640871164662917e-06,
+      "loss": 0.0689,
+      "step": 10297
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 8.64073787753579e-06,
+      "loss": 0.1191,
+      "step": 10298
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 8.640604590408658e-06,
+      "loss": 0.091,
+      "step": 10299
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 8.64047130328153e-06,
+      "loss": 0.0643,
+      "step": 10300
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 8.6403380161544e-06,
+      "loss": 0.0742,
+      "step": 10301
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 8.640204729027271e-06,
+      "loss": 0.0786,
+      "step": 10302
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 8.640071441900142e-06,
+      "loss": 0.1192,
+      "step": 10303
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 8.639938154773013e-06,
+      "loss": 0.0763,
+      "step": 10304
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 8.639804867645883e-06,
+      "loss": 0.1604,
+      "step": 10305
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 8.639671580518754e-06,
+      "loss": 0.0788,
+      "step": 10306
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 8.639538293391625e-06,
+      "loss": 0.0996,
+      "step": 10307
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 8.639405006264496e-06,
+      "loss": 0.138,
+      "step": 10308
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 8.639271719137367e-06,
+      "loss": 0.111,
+      "step": 10309
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 8.639138432010236e-06,
+      "loss": 0.1088,
+      "step": 10310
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 8.639005144883109e-06,
+      "loss": 0.1075,
+      "step": 10311
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 8.638871857755978e-06,
+      "loss": 0.1019,
+      "step": 10312
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 8.638738570628849e-06,
+      "loss": 0.091,
+      "step": 10313
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 8.63860528350172e-06,
+      "loss": 0.1174,
+      "step": 10314
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 8.638471996374591e-06,
+      "loss": 0.0855,
+      "step": 10315
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 8.638338709247462e-06,
+      "loss": 0.085,
+      "step": 10316
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 8.638205422120333e-06,
+      "loss": 0.0789,
+      "step": 10317
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 8.638072134993202e-06,
+      "loss": 0.1244,
+      "step": 10318
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 8.637938847866075e-06,
+      "loss": 0.1183,
+      "step": 10319
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 8.637805560738944e-06,
+      "loss": 0.0861,
+      "step": 10320
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 8.637672273611815e-06,
+      "loss": 0.0886,
+      "step": 10321
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 8.637538986484686e-06,
+      "loss": 0.0607,
+      "step": 10322
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 8.637405699357557e-06,
+      "loss": 0.1299,
+      "step": 10323
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 8.637272412230428e-06,
+      "loss": 0.11,
+      "step": 10324
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 8.6371391251033e-06,
+      "loss": 0.0729,
+      "step": 10325
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 8.637005837976169e-06,
+      "loss": 0.1471,
+      "step": 10326
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 8.63687255084904e-06,
+      "loss": 0.1069,
+      "step": 10327
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 8.63673926372191e-06,
+      "loss": 0.0955,
+      "step": 10328
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 8.636605976594782e-06,
+      "loss": 0.115,
+      "step": 10329
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 8.636472689467653e-06,
+      "loss": 0.1038,
+      "step": 10330
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 8.636339402340522e-06,
+      "loss": 0.1045,
+      "step": 10331
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 8.636206115213393e-06,
+      "loss": 0.0887,
+      "step": 10332
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 8.636072828086264e-06,
+      "loss": 0.0904,
+      "step": 10333
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 8.635939540959135e-06,
+      "loss": 0.129,
+      "step": 10334
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 8.635806253832006e-06,
+      "loss": 0.1034,
+      "step": 10335
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 8.635672966704877e-06,
+      "loss": 0.168,
+      "step": 10336
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 8.635539679577746e-06,
+      "loss": 0.1104,
+      "step": 10337
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 8.635406392450619e-06,
+      "loss": 0.0801,
+      "step": 10338
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 8.635273105323488e-06,
+      "loss": 0.1207,
+      "step": 10339
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 8.635139818196359e-06,
+      "loss": 0.1236,
+      "step": 10340
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 8.63500653106923e-06,
+      "loss": 0.0885,
+      "step": 10341
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 8.634873243942101e-06,
+      "loss": 0.1242,
+      "step": 10342
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 8.634739956814972e-06,
+      "loss": 0.0861,
+      "step": 10343
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 8.634606669687843e-06,
+      "loss": 0.1735,
+      "step": 10344
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 8.634473382560712e-06,
+      "loss": 0.1002,
+      "step": 10345
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 8.634340095433583e-06,
+      "loss": 0.0918,
+      "step": 10346
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 8.634206808306454e-06,
+      "loss": 0.1023,
+      "step": 10347
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 8.634073521179325e-06,
+      "loss": 0.1049,
+      "step": 10348
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 8.633940234052196e-06,
+      "loss": 0.1054,
+      "step": 10349
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 8.633806946925066e-06,
+      "loss": 0.1045,
+      "step": 10350
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 8.633673659797938e-06,
+      "loss": 0.0893,
+      "step": 10351
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 8.633540372670808e-06,
+      "loss": 0.1413,
+      "step": 10352
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 8.633407085543679e-06,
+      "loss": 0.1519,
+      "step": 10353
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 8.63327379841655e-06,
+      "loss": 0.0999,
+      "step": 10354
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 8.63314051128942e-06,
+      "loss": 0.1212,
+      "step": 10355
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 8.633007224162292e-06,
+      "loss": 0.1066,
+      "step": 10356
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 8.632873937035163e-06,
+      "loss": 0.113,
+      "step": 10357
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 8.632740649908032e-06,
+      "loss": 0.1002,
+      "step": 10358
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 8.632607362780903e-06,
+      "loss": 0.1262,
+      "step": 10359
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 8.632474075653774e-06,
+      "loss": 0.1322,
+      "step": 10360
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 8.632340788526645e-06,
+      "loss": 0.1209,
+      "step": 10361
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 8.632207501399516e-06,
+      "loss": 0.1331,
+      "step": 10362
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 8.632074214272387e-06,
+      "loss": 0.0967,
+      "step": 10363
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 8.631940927145256e-06,
+      "loss": 0.0783,
+      "step": 10364
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 8.631807640018127e-06,
+      "loss": 0.1131,
+      "step": 10365
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 8.631674352890998e-06,
+      "loss": 0.0902,
+      "step": 10366
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 8.63154106576387e-06,
+      "loss": 0.0915,
+      "step": 10367
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 8.63140777863674e-06,
+      "loss": 0.0558,
+      "step": 10368
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 8.63127449150961e-06,
+      "loss": 0.1153,
+      "step": 10369
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 8.631141204382482e-06,
+      "loss": 0.0831,
+      "step": 10370
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 8.631007917255351e-06,
+      "loss": 0.0848,
+      "step": 10371
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 8.630874630128222e-06,
+      "loss": 0.115,
+      "step": 10372
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 8.630741343001093e-06,
+      "loss": 0.1083,
+      "step": 10373
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 8.630608055873964e-06,
+      "loss": 0.1046,
+      "step": 10374
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 8.630474768746835e-06,
+      "loss": 0.1738,
+      "step": 10375
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 8.630341481619706e-06,
+      "loss": 0.1759,
+      "step": 10376
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 8.630208194492576e-06,
+      "loss": 0.1246,
+      "step": 10377
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 8.630074907365448e-06,
+      "loss": 0.0971,
+      "step": 10378
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 8.629941620238318e-06,
+      "loss": 0.1413,
+      "step": 10379
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 8.629808333111189e-06,
+      "loss": 0.1046,
+      "step": 10380
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 8.62967504598406e-06,
+      "loss": 0.1032,
+      "step": 10381
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 8.62954175885693e-06,
+      "loss": 0.1412,
+      "step": 10382
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 8.629408471729802e-06,
+      "loss": 0.0927,
+      "step": 10383
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 8.629275184602673e-06,
+      "loss": 0.0737,
+      "step": 10384
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 8.629141897475542e-06,
+      "loss": 0.1106,
+      "step": 10385
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 8.629008610348413e-06,
+      "loss": 0.0697,
+      "step": 10386
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 8.628875323221284e-06,
+      "loss": 0.1742,
+      "step": 10387
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 8.628742036094155e-06,
+      "loss": 0.096,
+      "step": 10388
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 8.628608748967026e-06,
+      "loss": 0.0897,
+      "step": 10389
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 8.628475461839895e-06,
+      "loss": 0.0874,
+      "step": 10390
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 8.628342174712766e-06,
+      "loss": 0.0764,
+      "step": 10391
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 8.628208887585637e-06,
+      "loss": 0.0947,
+      "step": 10392
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 8.628075600458508e-06,
+      "loss": 0.1047,
+      "step": 10393
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 8.62794231333138e-06,
+      "loss": 0.117,
+      "step": 10394
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 8.62780902620425e-06,
+      "loss": 0.123,
+      "step": 10395
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 8.62767573907712e-06,
+      "loss": 0.0804,
+      "step": 10396
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 8.627542451949992e-06,
+      "loss": 0.1447,
+      "step": 10397
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 8.627409164822862e-06,
+      "loss": 0.1114,
+      "step": 10398
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 8.627275877695733e-06,
+      "loss": 0.1229,
+      "step": 10399
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 8.627142590568604e-06,
+      "loss": 0.1647,
+      "step": 10400
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 8.627009303441475e-06,
+      "loss": 0.0842,
+      "step": 10401
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 8.626876016314346e-06,
+      "loss": 0.1305,
+      "step": 10402
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 8.626742729187217e-06,
+      "loss": 0.0835,
+      "step": 10403
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 8.626609442060086e-06,
+      "loss": 0.0942,
+      "step": 10404
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 8.626476154932957e-06,
+      "loss": 0.111,
+      "step": 10405
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 8.626342867805828e-06,
+      "loss": 0.1026,
+      "step": 10406
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 8.626209580678699e-06,
+      "loss": 0.0878,
+      "step": 10407
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 8.62607629355157e-06,
+      "loss": 0.133,
+      "step": 10408
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 8.625943006424439e-06,
+      "loss": 0.0926,
+      "step": 10409
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 8.625809719297312e-06,
+      "loss": 0.114,
+      "step": 10410
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 8.625676432170181e-06,
+      "loss": 0.109,
+      "step": 10411
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 8.625543145043052e-06,
+      "loss": 0.0818,
+      "step": 10412
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 8.625409857915923e-06,
+      "loss": 0.1063,
+      "step": 10413
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 8.625276570788794e-06,
+      "loss": 0.0995,
+      "step": 10414
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 8.625143283661665e-06,
+      "loss": 0.1144,
+      "step": 10415
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 8.625009996534536e-06,
+      "loss": 0.0947,
+      "step": 10416
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 8.624876709407405e-06,
+      "loss": 0.1132,
+      "step": 10417
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 8.624743422280278e-06,
+      "loss": 0.1145,
+      "step": 10418
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 8.624610135153147e-06,
+      "loss": 0.0887,
+      "step": 10419
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 8.624476848026018e-06,
+      "loss": 0.0875,
+      "step": 10420
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 8.62434356089889e-06,
+      "loss": 0.1507,
+      "step": 10421
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 8.62421027377176e-06,
+      "loss": 0.0801,
+      "step": 10422
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 8.624076986644631e-06,
+      "loss": 0.1157,
+      "step": 10423
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 8.6239436995175e-06,
+      "loss": 0.0705,
+      "step": 10424
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 8.623810412390372e-06,
+      "loss": 0.1092,
+      "step": 10425
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 8.623677125263243e-06,
+      "loss": 0.0721,
+      "step": 10426
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 8.623543838136114e-06,
+      "loss": 0.0956,
+      "step": 10427
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 8.623410551008985e-06,
+      "loss": 0.1294,
+      "step": 10428
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 8.623277263881856e-06,
+      "loss": 0.1317,
+      "step": 10429
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 8.623143976754725e-06,
+      "loss": 0.0937,
+      "step": 10430
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 8.623010689627596e-06,
+      "loss": 0.1324,
+      "step": 10431
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 8.622877402500467e-06,
+      "loss": 0.1276,
+      "step": 10432
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 8.622744115373338e-06,
+      "loss": 0.1263,
+      "step": 10433
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 8.622610828246209e-06,
+      "loss": 0.113,
+      "step": 10434
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 8.62247754111908e-06,
+      "loss": 0.0869,
+      "step": 10435
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 8.62234425399195e-06,
+      "loss": 0.0883,
+      "step": 10436
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 8.622210966864822e-06,
+      "loss": 0.1211,
+      "step": 10437
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 8.622077679737691e-06,
+      "loss": 0.0613,
+      "step": 10438
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 8.621944392610562e-06,
+      "loss": 0.0979,
+      "step": 10439
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 8.621811105483433e-06,
+      "loss": 0.0803,
+      "step": 10440
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 8.621677818356304e-06,
+      "loss": 0.1324,
+      "step": 10441
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 8.621544531229175e-06,
+      "loss": 0.1021,
+      "step": 10442
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 8.621411244102046e-06,
+      "loss": 0.089,
+      "step": 10443
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 8.621277956974915e-06,
+      "loss": 0.0882,
+      "step": 10444
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 8.621144669847786e-06,
+      "loss": 0.1065,
+      "step": 10445
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 8.621011382720657e-06,
+      "loss": 0.0716,
+      "step": 10446
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 8.620878095593528e-06,
+      "loss": 0.0856,
+      "step": 10447
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 8.6207448084664e-06,
+      "loss": 0.1023,
+      "step": 10448
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 8.620611521339269e-06,
+      "loss": 0.0804,
+      "step": 10449
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 8.620478234212141e-06,
+      "loss": 0.096,
+      "step": 10450
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 8.62034494708501e-06,
+      "loss": 0.126,
+      "step": 10451
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 8.620211659957882e-06,
+      "loss": 0.1313,
+      "step": 10452
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 8.620078372830753e-06,
+      "loss": 0.0948,
+      "step": 10453
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 8.619945085703624e-06,
+      "loss": 0.1028,
+      "step": 10454
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 8.619811798576495e-06,
+      "loss": 0.1095,
+      "step": 10455
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 8.619678511449366e-06,
+      "loss": 0.1117,
+      "step": 10456
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 8.619545224322235e-06,
+      "loss": 0.1128,
+      "step": 10457
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 8.619411937195106e-06,
+      "loss": 0.0817,
+      "step": 10458
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 8.619278650067977e-06,
+      "loss": 0.0973,
+      "step": 10459
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 8.619145362940848e-06,
+      "loss": 0.1039,
+      "step": 10460
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 8.619012075813719e-06,
+      "loss": 0.0912,
+      "step": 10461
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 8.61887878868659e-06,
+      "loss": 0.1343,
+      "step": 10462
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 8.61874550155946e-06,
+      "loss": 0.163,
+      "step": 10463
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 8.61861221443233e-06,
+      "loss": 0.128,
+      "step": 10464
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 8.618478927305201e-06,
+      "loss": 0.1946,
+      "step": 10465
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 8.618345640178072e-06,
+      "loss": 0.1053,
+      "step": 10466
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 8.618212353050943e-06,
+      "loss": 0.117,
+      "step": 10467
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 8.618079065923813e-06,
+      "loss": 0.0976,
+      "step": 10468
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 8.617945778796685e-06,
+      "loss": 0.099,
+      "step": 10469
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 8.617812491669555e-06,
+      "loss": 0.1153,
+      "step": 10470
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 8.617679204542426e-06,
+      "loss": 0.111,
+      "step": 10471
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 8.617545917415297e-06,
+      "loss": 0.0925,
+      "step": 10472
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 8.617412630288168e-06,
+      "loss": 0.1655,
+      "step": 10473
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 8.617279343161039e-06,
+      "loss": 0.1404,
+      "step": 10474
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 8.61714605603391e-06,
+      "loss": 0.1197,
+      "step": 10475
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 8.617012768906779e-06,
+      "loss": 0.0927,
+      "step": 10476
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 8.616879481779652e-06,
+      "loss": 0.1138,
+      "step": 10477
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 8.61674619465252e-06,
+      "loss": 0.1483,
+      "step": 10478
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 8.616612907525392e-06,
+      "loss": 0.0773,
+      "step": 10479
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 8.616479620398263e-06,
+      "loss": 0.1085,
+      "step": 10480
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 8.616346333271134e-06,
+      "loss": 0.1234,
+      "step": 10481
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 8.616213046144005e-06,
+      "loss": 0.0884,
+      "step": 10482
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 8.616079759016874e-06,
+      "loss": 0.1047,
+      "step": 10483
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 8.615946471889745e-06,
+      "loss": 0.1192,
+      "step": 10484
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 8.615813184762616e-06,
+      "loss": 0.0955,
+      "step": 10485
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 8.615679897635487e-06,
+      "loss": 0.103,
+      "step": 10486
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 8.615546610508358e-06,
+      "loss": 0.1098,
+      "step": 10487
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 8.615413323381229e-06,
+      "loss": 0.114,
+      "step": 10488
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 8.615280036254098e-06,
+      "loss": 0.0716,
+      "step": 10489
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 8.615146749126971e-06,
+      "loss": 0.1309,
+      "step": 10490
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 8.61501346199984e-06,
+      "loss": 0.0824,
+      "step": 10491
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 8.614880174872711e-06,
+      "loss": 0.1128,
+      "step": 10492
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 8.614746887745582e-06,
+      "loss": 0.101,
+      "step": 10493
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 8.614613600618453e-06,
+      "loss": 0.1021,
+      "step": 10494
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 8.614480313491324e-06,
+      "loss": 0.1189,
+      "step": 10495
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 8.614347026364195e-06,
+      "loss": 0.0979,
+      "step": 10496
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 8.614213739237065e-06,
+      "loss": 0.1092,
+      "step": 10497
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 8.614080452109936e-06,
+      "loss": 0.0946,
+      "step": 10498
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 8.613947164982807e-06,
+      "loss": 0.1207,
+      "step": 10499
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 8.613813877855678e-06,
+      "loss": 0.1446,
+      "step": 10500
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 8.613680590728549e-06,
+      "loss": 0.1329,
+      "step": 10501
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 8.613547303601418e-06,
+      "loss": 0.1147,
+      "step": 10502
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 8.613414016474289e-06,
+      "loss": 0.0969,
+      "step": 10503
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 8.61328072934716e-06,
+      "loss": 0.1115,
+      "step": 10504
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 8.613147442220031e-06,
+      "loss": 0.1191,
+      "step": 10505
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 8.613014155092902e-06,
+      "loss": 0.075,
+      "step": 10506
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 8.612880867965773e-06,
+      "loss": 0.0964,
+      "step": 10507
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 8.612747580838642e-06,
+      "loss": 0.1027,
+      "step": 10508
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 8.612614293711515e-06,
+      "loss": 0.0677,
+      "step": 10509
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 8.612481006584384e-06,
+      "loss": 0.0806,
+      "step": 10510
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 8.612347719457255e-06,
+      "loss": 0.0734,
+      "step": 10511
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 8.612214432330126e-06,
+      "loss": 0.1555,
+      "step": 10512
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 8.612081145202997e-06,
+      "loss": 0.1398,
+      "step": 10513
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 8.611947858075868e-06,
+      "loss": 0.1233,
+      "step": 10514
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 8.611814570948739e-06,
+      "loss": 0.1157,
+      "step": 10515
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 8.611681283821608e-06,
+      "loss": 0.0903,
+      "step": 10516
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 8.611547996694481e-06,
+      "loss": 0.1303,
+      "step": 10517
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 8.61141470956735e-06,
+      "loss": 0.083,
+      "step": 10518
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 8.611281422440221e-06,
+      "loss": 0.1286,
+      "step": 10519
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 8.611148135313092e-06,
+      "loss": 0.0712,
+      "step": 10520
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 8.611014848185963e-06,
+      "loss": 0.1031,
+      "step": 10521
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 8.610881561058834e-06,
+      "loss": 0.072,
+      "step": 10522
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 8.610748273931704e-06,
+      "loss": 0.1284,
+      "step": 10523
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 8.610614986804575e-06,
+      "loss": 0.0935,
+      "step": 10524
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 8.610481699677446e-06,
+      "loss": 0.078,
+      "step": 10525
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 8.610348412550317e-06,
+      "loss": 0.125,
+      "step": 10526
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 8.610215125423188e-06,
+      "loss": 0.1102,
+      "step": 10527
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 8.610081838296059e-06,
+      "loss": 0.1249,
+      "step": 10528
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 8.609948551168928e-06,
+      "loss": 0.1144,
+      "step": 10529
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 8.609815264041799e-06,
+      "loss": 0.0936,
+      "step": 10530
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 8.60968197691467e-06,
+      "loss": 0.0944,
+      "step": 10531
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 8.609548689787541e-06,
+      "loss": 0.1115,
+      "step": 10532
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 8.609415402660412e-06,
+      "loss": 0.0787,
+      "step": 10533
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 8.609282115533283e-06,
+      "loss": 0.1102,
+      "step": 10534
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 8.609148828406152e-06,
+      "loss": 0.1028,
+      "step": 10535
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 8.609015541279025e-06,
+      "loss": 0.1513,
+      "step": 10536
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 8.608882254151894e-06,
+      "loss": 0.0994,
+      "step": 10537
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 8.608748967024765e-06,
+      "loss": 0.0799,
+      "step": 10538
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 8.608615679897636e-06,
+      "loss": 0.0802,
+      "step": 10539
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 8.608482392770507e-06,
+      "loss": 0.0922,
+      "step": 10540
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 8.608349105643378e-06,
+      "loss": 0.1112,
+      "step": 10541
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 8.608215818516248e-06,
+      "loss": 0.1071,
+      "step": 10542
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 8.608082531389119e-06,
+      "loss": 0.1452,
+      "step": 10543
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 8.60794924426199e-06,
+      "loss": 0.1262,
+      "step": 10544
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 8.60781595713486e-06,
+      "loss": 0.0771,
+      "step": 10545
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 8.607682670007732e-06,
+      "loss": 0.092,
+      "step": 10546
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 8.607549382880603e-06,
+      "loss": 0.1214,
+      "step": 10547
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 8.607416095753472e-06,
+      "loss": 0.1364,
+      "step": 10548
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 8.607282808626344e-06,
+      "loss": 0.1279,
+      "step": 10549
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 8.607149521499214e-06,
+      "loss": 0.071,
+      "step": 10550
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 8.607016234372085e-06,
+      "loss": 0.1489,
+      "step": 10551
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 8.606882947244956e-06,
+      "loss": 0.1165,
+      "step": 10552
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 8.606749660117827e-06,
+      "loss": 0.0666,
+      "step": 10553
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 8.606616372990698e-06,
+      "loss": 0.1014,
+      "step": 10554
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 8.606483085863569e-06,
+      "loss": 0.1154,
+      "step": 10555
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 8.606349798736438e-06,
+      "loss": 0.1159,
+      "step": 10556
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 8.60621651160931e-06,
+      "loss": 0.138,
+      "step": 10557
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 8.60608322448218e-06,
+      "loss": 0.1321,
+      "step": 10558
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 8.605949937355051e-06,
+      "loss": 0.0776,
+      "step": 10559
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 8.605816650227922e-06,
+      "loss": 0.0893,
+      "step": 10560
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 8.605683363100791e-06,
+      "loss": 0.1096,
+      "step": 10561
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 8.605550075973664e-06,
+      "loss": 0.1229,
+      "step": 10562
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 8.605416788846533e-06,
+      "loss": 0.1046,
+      "step": 10563
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 8.605283501719404e-06,
+      "loss": 0.1171,
+      "step": 10564
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 8.605150214592275e-06,
+      "loss": 0.0889,
+      "step": 10565
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 8.605016927465146e-06,
+      "loss": 0.1137,
+      "step": 10566
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 8.604883640338017e-06,
+      "loss": 0.0719,
+      "step": 10567
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 8.604750353210888e-06,
+      "loss": 0.0896,
+      "step": 10568
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 8.604617066083758e-06,
+      "loss": 0.1214,
+      "step": 10569
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 8.604483778956629e-06,
+      "loss": 0.1246,
+      "step": 10570
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 8.6043504918295e-06,
+      "loss": 0.1041,
+      "step": 10571
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 8.60421720470237e-06,
+      "loss": 0.1262,
+      "step": 10572
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 8.604083917575242e-06,
+      "loss": 0.1186,
+      "step": 10573
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 8.603950630448113e-06,
+      "loss": 0.1275,
+      "step": 10574
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 8.603817343320982e-06,
+      "loss": 0.1241,
+      "step": 10575
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 8.603684056193855e-06,
+      "loss": 0.1067,
+      "step": 10576
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 8.603550769066724e-06,
+      "loss": 0.0999,
+      "step": 10577
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 8.603417481939595e-06,
+      "loss": 0.0882,
+      "step": 10578
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 8.603284194812466e-06,
+      "loss": 0.1415,
+      "step": 10579
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 8.603150907685337e-06,
+      "loss": 0.1195,
+      "step": 10580
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 8.603017620558208e-06,
+      "loss": 0.0969,
+      "step": 10581
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 8.602884333431077e-06,
+      "loss": 0.0916,
+      "step": 10582
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 8.602751046303948e-06,
+      "loss": 0.099,
+      "step": 10583
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 8.602617759176819e-06,
+      "loss": 0.1043,
+      "step": 10584
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 8.60248447204969e-06,
+      "loss": 0.0949,
+      "step": 10585
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 8.602351184922561e-06,
+      "loss": 0.1756,
+      "step": 10586
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 8.602217897795432e-06,
+      "loss": 0.0955,
+      "step": 10587
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 8.602084610668301e-06,
+      "loss": 0.102,
+      "step": 10588
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 8.601951323541174e-06,
+      "loss": 0.1248,
+      "step": 10589
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 8.601818036414043e-06,
+      "loss": 0.105,
+      "step": 10590
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 8.601684749286914e-06,
+      "loss": 0.1101,
+      "step": 10591
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 8.601551462159785e-06,
+      "loss": 0.1456,
+      "step": 10592
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 8.601418175032656e-06,
+      "loss": 0.0929,
+      "step": 10593
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 8.601284887905527e-06,
+      "loss": 0.1061,
+      "step": 10594
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 8.601151600778398e-06,
+      "loss": 0.1278,
+      "step": 10595
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 8.601018313651268e-06,
+      "loss": 0.0903,
+      "step": 10596
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 8.600885026524139e-06,
+      "loss": 0.139,
+      "step": 10597
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 8.60075173939701e-06,
+      "loss": 0.0956,
+      "step": 10598
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 8.60061845226988e-06,
+      "loss": 0.0863,
+      "step": 10599
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 8.600485165142752e-06,
+      "loss": 0.1137,
+      "step": 10600
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 8.600351878015621e-06,
+      "loss": 0.107,
+      "step": 10601
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 8.600218590888492e-06,
+      "loss": 0.1154,
+      "step": 10602
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 8.600085303761363e-06,
+      "loss": 0.0899,
+      "step": 10603
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 8.599952016634234e-06,
+      "loss": 0.115,
+      "step": 10604
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 8.599818729507105e-06,
+      "loss": 0.1463,
+      "step": 10605
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 8.599685442379976e-06,
+      "loss": 0.1388,
+      "step": 10606
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 8.599552155252845e-06,
+      "loss": 0.1113,
+      "step": 10607
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 8.599418868125718e-06,
+      "loss": 0.0912,
+      "step": 10608
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 8.599285580998587e-06,
+      "loss": 0.104,
+      "step": 10609
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 8.599152293871458e-06,
+      "loss": 0.0934,
+      "step": 10610
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 8.59901900674433e-06,
+      "loss": 0.0752,
+      "step": 10611
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 8.5988857196172e-06,
+      "loss": 0.1292,
+      "step": 10612
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 8.598752432490071e-06,
+      "loss": 0.1143,
+      "step": 10613
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 8.598619145362942e-06,
+      "loss": 0.1063,
+      "step": 10614
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 8.598485858235812e-06,
+      "loss": 0.1057,
+      "step": 10615
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 8.598352571108684e-06,
+      "loss": 0.0863,
+      "step": 10616
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 8.598219283981553e-06,
+      "loss": 0.0953,
+      "step": 10617
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 8.598085996854424e-06,
+      "loss": 0.0831,
+      "step": 10618
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 8.597952709727295e-06,
+      "loss": 0.0825,
+      "step": 10619
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 8.597819422600165e-06,
+      "loss": 0.1029,
+      "step": 10620
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 8.597686135473037e-06,
+      "loss": 0.0843,
+      "step": 10621
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 8.597552848345907e-06,
+      "loss": 0.1464,
+      "step": 10622
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 8.597419561218778e-06,
+      "loss": 0.1037,
+      "step": 10623
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 8.597286274091649e-06,
+      "loss": 0.104,
+      "step": 10624
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 8.59715298696452e-06,
+      "loss": 0.0992,
+      "step": 10625
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 8.59701969983739e-06,
+      "loss": 0.1266,
+      "step": 10626
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 8.596886412710262e-06,
+      "loss": 0.1202,
+      "step": 10627
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 8.596753125583131e-06,
+      "loss": 0.0969,
+      "step": 10628
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 8.596619838456004e-06,
+      "loss": 0.1001,
+      "step": 10629
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 8.596486551328873e-06,
+      "loss": 0.1029,
+      "step": 10630
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 8.596353264201744e-06,
+      "loss": 0.113,
+      "step": 10631
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 8.596219977074615e-06,
+      "loss": 0.0767,
+      "step": 10632
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 8.596086689947486e-06,
+      "loss": 0.1339,
+      "step": 10633
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 8.595953402820355e-06,
+      "loss": 0.0929,
+      "step": 10634
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 8.595820115693228e-06,
+      "loss": 0.1153,
+      "step": 10635
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 8.595686828566097e-06,
+      "loss": 0.0852,
+      "step": 10636
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 8.595553541438968e-06,
+      "loss": 0.1061,
+      "step": 10637
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 8.59542025431184e-06,
+      "loss": 0.1467,
+      "step": 10638
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 8.595286967184709e-06,
+      "loss": 0.1248,
+      "step": 10639
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 8.595153680057581e-06,
+      "loss": 0.0764,
+      "step": 10640
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 8.59502039293045e-06,
+      "loss": 0.1517,
+      "step": 10641
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 8.594887105803322e-06,
+      "loss": 0.1275,
+      "step": 10642
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 8.594753818676193e-06,
+      "loss": 0.0928,
+      "step": 10643
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 8.594620531549064e-06,
+      "loss": 0.1102,
+      "step": 10644
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 8.594487244421935e-06,
+      "loss": 0.1515,
+      "step": 10645
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 8.594353957294806e-06,
+      "loss": 0.1384,
+      "step": 10646
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 8.594220670167675e-06,
+      "loss": 0.0844,
+      "step": 10647
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 8.594087383040548e-06,
+      "loss": 0.087,
+      "step": 10648
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 8.593954095913417e-06,
+      "loss": 0.0953,
+      "step": 10649
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 8.593820808786288e-06,
+      "loss": 0.0959,
+      "step": 10650
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 8.593687521659159e-06,
+      "loss": 0.1288,
+      "step": 10651
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 8.59355423453203e-06,
+      "loss": 0.1026,
+      "step": 10652
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 8.5934209474049e-06,
+      "loss": 0.0747,
+      "step": 10653
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 8.593287660277772e-06,
+      "loss": 0.0856,
+      "step": 10654
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 8.593154373150641e-06,
+      "loss": 0.1085,
+      "step": 10655
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 8.593021086023514e-06,
+      "loss": 0.1268,
+      "step": 10656
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 8.592887798896383e-06,
+      "loss": 0.0666,
+      "step": 10657
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 8.592754511769254e-06,
+      "loss": 0.1078,
+      "step": 10658
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 8.592621224642125e-06,
+      "loss": 0.123,
+      "step": 10659
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 8.592487937514994e-06,
+      "loss": 0.0909,
+      "step": 10660
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 8.592354650387867e-06,
+      "loss": 0.1206,
+      "step": 10661
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 8.592221363260736e-06,
+      "loss": 0.1289,
+      "step": 10662
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 8.592088076133607e-06,
+      "loss": 0.0847,
+      "step": 10663
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 8.591954789006478e-06,
+      "loss": 0.0896,
+      "step": 10664
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 8.59182150187935e-06,
+      "loss": 0.1057,
+      "step": 10665
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 8.59168821475222e-06,
+      "loss": 0.1055,
+      "step": 10666
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 8.591554927625091e-06,
+      "loss": 0.1005,
+      "step": 10667
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 8.59142164049796e-06,
+      "loss": 0.1243,
+      "step": 10668
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 8.591288353370832e-06,
+      "loss": 0.1255,
+      "step": 10669
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 8.591155066243703e-06,
+      "loss": 0.0755,
+      "step": 10670
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 8.591021779116574e-06,
+      "loss": 0.119,
+      "step": 10671
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 8.590888491989445e-06,
+      "loss": 0.0936,
+      "step": 10672
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 8.590755204862316e-06,
+      "loss": 0.1109,
+      "step": 10673
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 8.590621917735185e-06,
+      "loss": 0.0865,
+      "step": 10674
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 8.590488630608058e-06,
+      "loss": 0.0963,
+      "step": 10675
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 8.590355343480927e-06,
+      "loss": 0.0857,
+      "step": 10676
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 8.590222056353798e-06,
+      "loss": 0.0909,
+      "step": 10677
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 8.590088769226669e-06,
+      "loss": 0.1082,
+      "step": 10678
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 8.589955482099538e-06,
+      "loss": 0.1222,
+      "step": 10679
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 8.589822194972411e-06,
+      "loss": 0.1006,
+      "step": 10680
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 8.58968890784528e-06,
+      "loss": 0.1145,
+      "step": 10681
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 8.589555620718151e-06,
+      "loss": 0.1,
+      "step": 10682
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 8.589422333591022e-06,
+      "loss": 0.1118,
+      "step": 10683
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 8.589289046463893e-06,
+      "loss": 0.1121,
+      "step": 10684
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 8.589155759336764e-06,
+      "loss": 0.1121,
+      "step": 10685
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 8.589022472209635e-06,
+      "loss": 0.1796,
+      "step": 10686
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 8.588889185082504e-06,
+      "loss": 0.0939,
+      "step": 10687
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 8.588755897955377e-06,
+      "loss": 0.0738,
+      "step": 10688
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 8.588622610828246e-06,
+      "loss": 0.1194,
+      "step": 10689
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 8.588489323701117e-06,
+      "loss": 0.0886,
+      "step": 10690
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 8.588356036573988e-06,
+      "loss": 0.0998,
+      "step": 10691
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 8.58822274944686e-06,
+      "loss": 0.1038,
+      "step": 10692
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 8.58808946231973e-06,
+      "loss": 0.0945,
+      "step": 10693
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 8.587956175192601e-06,
+      "loss": 0.1195,
+      "step": 10694
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 8.58782288806547e-06,
+      "loss": 0.1048,
+      "step": 10695
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 8.587689600938342e-06,
+      "loss": 0.1225,
+      "step": 10696
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 8.587556313811213e-06,
+      "loss": 0.1285,
+      "step": 10697
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 8.587423026684084e-06,
+      "loss": 0.1122,
+      "step": 10698
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 8.587289739556955e-06,
+      "loss": 0.1407,
+      "step": 10699
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 8.587156452429824e-06,
+      "loss": 0.0877,
+      "step": 10700
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 8.587023165302695e-06,
+      "loss": 0.1027,
+      "step": 10701
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 8.586889878175566e-06,
+      "loss": 0.076,
+      "step": 10702
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 8.586756591048437e-06,
+      "loss": 0.0827,
+      "step": 10703
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 8.586623303921308e-06,
+      "loss": 0.1031,
+      "step": 10704
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 8.586490016794179e-06,
+      "loss": 0.0781,
+      "step": 10705
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 8.586356729667048e-06,
+      "loss": 0.0733,
+      "step": 10706
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 8.586223442539921e-06,
+      "loss": 0.0772,
+      "step": 10707
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 8.58609015541279e-06,
+      "loss": 0.0951,
+      "step": 10708
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 8.585956868285661e-06,
+      "loss": 0.1171,
+      "step": 10709
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 8.585823581158532e-06,
+      "loss": 0.127,
+      "step": 10710
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 8.585690294031403e-06,
+      "loss": 0.1082,
+      "step": 10711
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 8.585557006904274e-06,
+      "loss": 0.1162,
+      "step": 10712
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 8.585423719777145e-06,
+      "loss": 0.1381,
+      "step": 10713
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 8.585290432650015e-06,
+      "loss": 0.0721,
+      "step": 10714
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 8.585157145522887e-06,
+      "loss": 0.0933,
+      "step": 10715
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 8.585023858395757e-06,
+      "loss": 0.097,
+      "step": 10716
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 8.584890571268628e-06,
+      "loss": 0.1308,
+      "step": 10717
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 8.584757284141499e-06,
+      "loss": 0.1166,
+      "step": 10718
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 8.584623997014368e-06,
+      "loss": 0.097,
+      "step": 10719
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 8.58449070988724e-06,
+      "loss": 0.1397,
+      "step": 10720
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 8.58435742276011e-06,
+      "loss": 0.0966,
+      "step": 10721
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 8.58422413563298e-06,
+      "loss": 0.1182,
+      "step": 10722
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 8.584090848505852e-06,
+      "loss": 0.1109,
+      "step": 10723
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 8.583957561378723e-06,
+      "loss": 0.1143,
+      "step": 10724
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 8.583824274251594e-06,
+      "loss": 0.0921,
+      "step": 10725
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 8.583690987124465e-06,
+      "loss": 0.1265,
+      "step": 10726
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 8.583557699997334e-06,
+      "loss": 0.1124,
+      "step": 10727
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 8.583424412870207e-06,
+      "loss": 0.1518,
+      "step": 10728
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 8.583291125743076e-06,
+      "loss": 0.0982,
+      "step": 10729
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 8.583157838615947e-06,
+      "loss": 0.0823,
+      "step": 10730
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 8.583024551488818e-06,
+      "loss": 0.0804,
+      "step": 10731
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 8.582891264361689e-06,
+      "loss": 0.1466,
+      "step": 10732
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 8.58275797723456e-06,
+      "loss": 0.1158,
+      "step": 10733
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 8.582624690107431e-06,
+      "loss": 0.1114,
+      "step": 10734
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 8.5824914029803e-06,
+      "loss": 0.1395,
+      "step": 10735
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 8.582358115853171e-06,
+      "loss": 0.0945,
+      "step": 10736
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 8.582224828726042e-06,
+      "loss": 0.0999,
+      "step": 10737
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 8.582091541598913e-06,
+      "loss": 0.1216,
+      "step": 10738
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 8.581958254471784e-06,
+      "loss": 0.0987,
+      "step": 10739
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 8.581824967344654e-06,
+      "loss": 0.1181,
+      "step": 10740
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 8.581691680217525e-06,
+      "loss": 0.1114,
+      "step": 10741
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 8.581558393090396e-06,
+      "loss": 0.0779,
+      "step": 10742
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 8.581425105963267e-06,
+      "loss": 0.1024,
+      "step": 10743
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 8.581291818836138e-06,
+      "loss": 0.0984,
+      "step": 10744
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 8.581158531709009e-06,
+      "loss": 0.0819,
+      "step": 10745
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 8.581025244581878e-06,
+      "loss": 0.0804,
+      "step": 10746
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 8.58089195745475e-06,
+      "loss": 0.0919,
+      "step": 10747
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 8.58075867032762e-06,
+      "loss": 0.1266,
+      "step": 10748
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 8.580625383200491e-06,
+      "loss": 0.1133,
+      "step": 10749
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 8.580492096073362e-06,
+      "loss": 0.0943,
+      "step": 10750
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 8.580358808946233e-06,
+      "loss": 0.0913,
+      "step": 10751
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 8.580225521819104e-06,
+      "loss": 0.0923,
+      "step": 10752
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 8.580092234691975e-06,
+      "loss": 0.0736,
+      "step": 10753
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 8.579958947564844e-06,
+      "loss": 0.0895,
+      "step": 10754
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 8.579825660437717e-06,
+      "loss": 0.0983,
+      "step": 10755
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 8.579692373310586e-06,
+      "loss": 0.0978,
+      "step": 10756
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 8.579559086183457e-06,
+      "loss": 0.138,
+      "step": 10757
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 8.579425799056328e-06,
+      "loss": 0.1224,
+      "step": 10758
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 8.579292511929197e-06,
+      "loss": 0.1149,
+      "step": 10759
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 8.57915922480207e-06,
+      "loss": 0.081,
+      "step": 10760
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 8.57902593767494e-06,
+      "loss": 0.1572,
+      "step": 10761
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 8.57889265054781e-06,
+      "loss": 0.1099,
+      "step": 10762
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 8.578759363420681e-06,
+      "loss": 0.1227,
+      "step": 10763
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 8.578626076293552e-06,
+      "loss": 0.1064,
+      "step": 10764
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 8.578492789166423e-06,
+      "loss": 0.0974,
+      "step": 10765
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 8.578359502039294e-06,
+      "loss": 0.1048,
+      "step": 10766
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 8.578226214912164e-06,
+      "loss": 0.1121,
+      "step": 10767
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 8.578092927785035e-06,
+      "loss": 0.1119,
+      "step": 10768
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 8.577959640657906e-06,
+      "loss": 0.1113,
+      "step": 10769
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 8.577826353530777e-06,
+      "loss": 0.1059,
+      "step": 10770
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 8.577693066403648e-06,
+      "loss": 0.0852,
+      "step": 10771
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 8.577559779276519e-06,
+      "loss": 0.1638,
+      "step": 10772
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 8.577426492149388e-06,
+      "loss": 0.0703,
+      "step": 10773
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 8.57729320502226e-06,
+      "loss": 0.082,
+      "step": 10774
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 8.57715991789513e-06,
+      "loss": 0.0745,
+      "step": 10775
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 8.577026630768001e-06,
+      "loss": 0.089,
+      "step": 10776
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 8.576893343640872e-06,
+      "loss": 0.0716,
+      "step": 10777
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 8.576760056513741e-06,
+      "loss": 0.1096,
+      "step": 10778
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 8.576626769386614e-06,
+      "loss": 0.122,
+      "step": 10779
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 8.576493482259483e-06,
+      "loss": 0.0775,
+      "step": 10780
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 8.576360195132354e-06,
+      "loss": 0.1027,
+      "step": 10781
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 8.576226908005225e-06,
+      "loss": 0.0967,
+      "step": 10782
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 8.576093620878096e-06,
+      "loss": 0.1117,
+      "step": 10783
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 8.575960333750967e-06,
+      "loss": 0.1125,
+      "step": 10784
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 8.575827046623838e-06,
+      "loss": 0.0514,
+      "step": 10785
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 8.575693759496708e-06,
+      "loss": 0.1029,
+      "step": 10786
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 8.57556047236958e-06,
+      "loss": 0.1021,
+      "step": 10787
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 8.57542718524245e-06,
+      "loss": 0.1028,
+      "step": 10788
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 8.57529389811532e-06,
+      "loss": 0.0976,
+      "step": 10789
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 8.575160610988192e-06,
+      "loss": 0.1283,
+      "step": 10790
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 8.575027323861063e-06,
+      "loss": 0.099,
+      "step": 10791
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 8.574894036733934e-06,
+      "loss": 0.0964,
+      "step": 10792
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 8.574760749606805e-06,
+      "loss": 0.0989,
+      "step": 10793
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 8.574627462479674e-06,
+      "loss": 0.1233,
+      "step": 10794
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 8.574494175352547e-06,
+      "loss": 0.1014,
+      "step": 10795
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 8.574360888225416e-06,
+      "loss": 0.1299,
+      "step": 10796
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 8.574227601098287e-06,
+      "loss": 0.1475,
+      "step": 10797
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 8.574094313971158e-06,
+      "loss": 0.1216,
+      "step": 10798
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 8.573961026844027e-06,
+      "loss": 0.1212,
+      "step": 10799
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 8.5738277397169e-06,
+      "loss": 0.1057,
+      "step": 10800
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 8.573694452589769e-06,
+      "loss": 0.0954,
+      "step": 10801
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 8.57356116546264e-06,
+      "loss": 0.1024,
+      "step": 10802
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 8.573427878335511e-06,
+      "loss": 0.1115,
+      "step": 10803
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 8.573294591208382e-06,
+      "loss": 0.0733,
+      "step": 10804
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 8.573161304081253e-06,
+      "loss": 0.0892,
+      "step": 10805
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 8.573028016954124e-06,
+      "loss": 0.1325,
+      "step": 10806
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 8.572894729826993e-06,
+      "loss": 0.104,
+      "step": 10807
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 8.572761442699864e-06,
+      "loss": 0.1195,
+      "step": 10808
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 8.572628155572735e-06,
+      "loss": 0.0885,
+      "step": 10809
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 8.572494868445606e-06,
+      "loss": 0.0866,
+      "step": 10810
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 8.572361581318477e-06,
+      "loss": 0.0904,
+      "step": 10811
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 8.572228294191348e-06,
+      "loss": 0.1048,
+      "step": 10812
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 8.572095007064218e-06,
+      "loss": 0.0734,
+      "step": 10813
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 8.57196171993709e-06,
+      "loss": 0.1605,
+      "step": 10814
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 8.57182843280996e-06,
+      "loss": 0.0914,
+      "step": 10815
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 8.57169514568283e-06,
+      "loss": 0.106,
+      "step": 10816
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 8.571561858555702e-06,
+      "loss": 0.1277,
+      "step": 10817
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 8.571428571428571e-06,
+      "loss": 0.071,
+      "step": 10818
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 8.571295284301444e-06,
+      "loss": 0.118,
+      "step": 10819
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 8.571161997174313e-06,
+      "loss": 0.0768,
+      "step": 10820
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 8.571028710047184e-06,
+      "loss": 0.126,
+      "step": 10821
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 8.570895422920055e-06,
+      "loss": 0.104,
+      "step": 10822
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 8.570762135792926e-06,
+      "loss": 0.152,
+      "step": 10823
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 8.570628848665797e-06,
+      "loss": 0.0913,
+      "step": 10824
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 8.570495561538668e-06,
+      "loss": 0.1207,
+      "step": 10825
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 8.570362274411537e-06,
+      "loss": 0.1055,
+      "step": 10826
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 8.57022898728441e-06,
+      "loss": 0.1165,
+      "step": 10827
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 8.57009570015728e-06,
+      "loss": 0.1328,
+      "step": 10828
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 8.56996241303015e-06,
+      "loss": 0.0803,
+      "step": 10829
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 8.569829125903021e-06,
+      "loss": 0.0721,
+      "step": 10830
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 8.569695838775892e-06,
+      "loss": 0.1198,
+      "step": 10831
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 8.569562551648763e-06,
+      "loss": 0.1471,
+      "step": 10832
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 8.569429264521634e-06,
+      "loss": 0.0776,
+      "step": 10833
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 8.569295977394503e-06,
+      "loss": 0.0646,
+      "step": 10834
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 8.569162690267374e-06,
+      "loss": 0.1187,
+      "step": 10835
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 8.569029403140245e-06,
+      "loss": 0.1153,
+      "step": 10836
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 8.568896116013116e-06,
+      "loss": 0.121,
+      "step": 10837
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 8.568762828885987e-06,
+      "loss": 0.083,
+      "step": 10838
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 8.568629541758857e-06,
+      "loss": 0.1108,
+      "step": 10839
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 8.568496254631728e-06,
+      "loss": 0.1236,
+      "step": 10840
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 8.568362967504599e-06,
+      "loss": 0.1185,
+      "step": 10841
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 8.56822968037747e-06,
+      "loss": 0.114,
+      "step": 10842
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 8.56809639325034e-06,
+      "loss": 0.097,
+      "step": 10843
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 8.567963106123212e-06,
+      "loss": 0.0794,
+      "step": 10844
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 8.567829818996081e-06,
+      "loss": 0.0961,
+      "step": 10845
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 8.567696531868954e-06,
+      "loss": 0.1255,
+      "step": 10846
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 8.567563244741823e-06,
+      "loss": 0.092,
+      "step": 10847
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 8.567429957614694e-06,
+      "loss": 0.0965,
+      "step": 10848
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 8.567296670487565e-06,
+      "loss": 0.108,
+      "step": 10849
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 8.567163383360436e-06,
+      "loss": 0.1421,
+      "step": 10850
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 8.567030096233307e-06,
+      "loss": 0.1343,
+      "step": 10851
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 8.566896809106178e-06,
+      "loss": 0.0825,
+      "step": 10852
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 8.566763521979047e-06,
+      "loss": 0.0824,
+      "step": 10853
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 8.56663023485192e-06,
+      "loss": 0.1434,
+      "step": 10854
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 8.56649694772479e-06,
+      "loss": 0.1274,
+      "step": 10855
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 8.56636366059766e-06,
+      "loss": 0.1082,
+      "step": 10856
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 8.566230373470531e-06,
+      "loss": 0.125,
+      "step": 10857
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 8.5660970863434e-06,
+      "loss": 0.0991,
+      "step": 10858
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 8.565963799216273e-06,
+      "loss": 0.1034,
+      "step": 10859
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 8.565830512089143e-06,
+      "loss": 0.0568,
+      "step": 10860
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 8.565697224962014e-06,
+      "loss": 0.0492,
+      "step": 10861
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 8.565563937834885e-06,
+      "loss": 0.0997,
+      "step": 10862
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 8.565430650707756e-06,
+      "loss": 0.111,
+      "step": 10863
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 8.565297363580627e-06,
+      "loss": 0.0776,
+      "step": 10864
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 8.565164076453498e-06,
+      "loss": 0.1073,
+      "step": 10865
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 8.565030789326367e-06,
+      "loss": 0.11,
+      "step": 10866
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 8.56489750219924e-06,
+      "loss": 0.1308,
+      "step": 10867
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 8.564764215072109e-06,
+      "loss": 0.1055,
+      "step": 10868
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 8.56463092794498e-06,
+      "loss": 0.0881,
+      "step": 10869
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 8.56449764081785e-06,
+      "loss": 0.0804,
+      "step": 10870
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 8.564364353690722e-06,
+      "loss": 0.0945,
+      "step": 10871
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 8.564231066563593e-06,
+      "loss": 0.1252,
+      "step": 10872
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 8.564097779436464e-06,
+      "loss": 0.1202,
+      "step": 10873
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 8.563964492309333e-06,
+      "loss": 0.1037,
+      "step": 10874
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 8.563831205182204e-06,
+      "loss": 0.1009,
+      "step": 10875
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 8.563697918055075e-06,
+      "loss": 0.104,
+      "step": 10876
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 8.563564630927946e-06,
+      "loss": 0.0957,
+      "step": 10877
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 8.563431343800817e-06,
+      "loss": 0.0762,
+      "step": 10878
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 8.563298056673686e-06,
+      "loss": 0.0768,
+      "step": 10879
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 8.563164769546557e-06,
+      "loss": 0.0747,
+      "step": 10880
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 8.563031482419428e-06,
+      "loss": 0.1058,
+      "step": 10881
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 8.5628981952923e-06,
+      "loss": 0.126,
+      "step": 10882
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 8.56276490816517e-06,
+      "loss": 0.0742,
+      "step": 10883
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 8.562631621038041e-06,
+      "loss": 0.1286,
+      "step": 10884
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 8.56249833391091e-06,
+      "loss": 0.1003,
+      "step": 10885
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 8.562365046783783e-06,
+      "loss": 0.1281,
+      "step": 10886
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 8.562231759656653e-06,
+      "loss": 0.1284,
+      "step": 10887
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 8.562098472529524e-06,
+      "loss": 0.1161,
+      "step": 10888
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 8.561965185402395e-06,
+      "loss": 0.11,
+      "step": 10889
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 8.561831898275266e-06,
+      "loss": 0.095,
+      "step": 10890
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 8.561698611148137e-06,
+      "loss": 0.1221,
+      "step": 10891
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 8.561565324021008e-06,
+      "loss": 0.1015,
+      "step": 10892
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 8.561432036893877e-06,
+      "loss": 0.1091,
+      "step": 10893
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 8.561298749766748e-06,
+      "loss": 0.1158,
+      "step": 10894
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 8.561165462639619e-06,
+      "loss": 0.1017,
+      "step": 10895
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 8.56103217551249e-06,
+      "loss": 0.0971,
+      "step": 10896
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 8.560898888385361e-06,
+      "loss": 0.1022,
+      "step": 10897
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 8.56076560125823e-06,
+      "loss": 0.1028,
+      "step": 10898
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 8.560632314131103e-06,
+      "loss": 0.0866,
+      "step": 10899
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 8.560499027003972e-06,
+      "loss": 0.1223,
+      "step": 10900
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 8.560365739876843e-06,
+      "loss": 0.1078,
+      "step": 10901
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 8.560232452749714e-06,
+      "loss": 0.1271,
+      "step": 10902
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 8.560099165622585e-06,
+      "loss": 0.1156,
+      "step": 10903
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 8.559965878495456e-06,
+      "loss": 0.1095,
+      "step": 10904
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 8.559832591368327e-06,
+      "loss": 0.0687,
+      "step": 10905
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 8.559699304241196e-06,
+      "loss": 0.0827,
+      "step": 10906
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 8.559566017114067e-06,
+      "loss": 0.1238,
+      "step": 10907
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 8.559432729986938e-06,
+      "loss": 0.0659,
+      "step": 10908
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 8.55929944285981e-06,
+      "loss": 0.1207,
+      "step": 10909
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 8.55916615573268e-06,
+      "loss": 0.0922,
+      "step": 10910
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 8.559032868605551e-06,
+      "loss": 0.0785,
+      "step": 10911
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 8.55889958147842e-06,
+      "loss": 0.1151,
+      "step": 10912
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 8.558766294351293e-06,
+      "loss": 0.151,
+      "step": 10913
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 8.558633007224163e-06,
+      "loss": 0.1119,
+      "step": 10914
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 8.558499720097034e-06,
+      "loss": 0.1069,
+      "step": 10915
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 8.558366432969905e-06,
+      "loss": 0.0685,
+      "step": 10916
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 8.558233145842774e-06,
+      "loss": 0.0969,
+      "step": 10917
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 8.558099858715647e-06,
+      "loss": 0.1042,
+      "step": 10918
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 8.557966571588516e-06,
+      "loss": 0.0641,
+      "step": 10919
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 8.557833284461387e-06,
+      "loss": 0.0819,
+      "step": 10920
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 8.557699997334258e-06,
+      "loss": 0.1246,
+      "step": 10921
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 8.557566710207129e-06,
+      "loss": 0.0861,
+      "step": 10922
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 8.55743342308e-06,
+      "loss": 0.0871,
+      "step": 10923
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 8.557300135952871e-06,
+      "loss": 0.1502,
+      "step": 10924
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 8.55716684882574e-06,
+      "loss": 0.1132,
+      "step": 10925
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 8.557033561698613e-06,
+      "loss": 0.0997,
+      "step": 10926
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 8.556900274571482e-06,
+      "loss": 0.0589,
+      "step": 10927
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 8.556766987444353e-06,
+      "loss": 0.0894,
+      "step": 10928
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 8.556633700317224e-06,
+      "loss": 0.1284,
+      "step": 10929
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 8.556500413190095e-06,
+      "loss": 0.0962,
+      "step": 10930
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 8.556367126062966e-06,
+      "loss": 0.0901,
+      "step": 10931
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 8.556233838935837e-06,
+      "loss": 0.1435,
+      "step": 10932
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 8.556100551808707e-06,
+      "loss": 0.0983,
+      "step": 10933
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 8.555967264681578e-06,
+      "loss": 0.1075,
+      "step": 10934
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 8.555833977554449e-06,
+      "loss": 0.099,
+      "step": 10935
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 8.55570069042732e-06,
+      "loss": 0.0847,
+      "step": 10936
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 8.55556740330019e-06,
+      "loss": 0.1332,
+      "step": 10937
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 8.55543411617306e-06,
+      "loss": 0.1531,
+      "step": 10938
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 8.55530082904593e-06,
+      "loss": 0.1013,
+      "step": 10939
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 8.555167541918802e-06,
+      "loss": 0.1223,
+      "step": 10940
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 8.555034254791673e-06,
+      "loss": 0.0809,
+      "step": 10941
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 8.554900967664544e-06,
+      "loss": 0.1266,
+      "step": 10942
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 8.554767680537415e-06,
+      "loss": 0.0939,
+      "step": 10943
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 8.554634393410284e-06,
+      "loss": 0.0804,
+      "step": 10944
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 8.554501106283157e-06,
+      "loss": 0.0941,
+      "step": 10945
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 8.554367819156026e-06,
+      "loss": 0.0995,
+      "step": 10946
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 8.554234532028897e-06,
+      "loss": 0.1384,
+      "step": 10947
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 8.554101244901768e-06,
+      "loss": 0.086,
+      "step": 10948
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 8.553967957774639e-06,
+      "loss": 0.1002,
+      "step": 10949
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 8.55383467064751e-06,
+      "loss": 0.066,
+      "step": 10950
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 8.553701383520381e-06,
+      "loss": 0.1024,
+      "step": 10951
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 8.55356809639325e-06,
+      "loss": 0.1172,
+      "step": 10952
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 8.553434809266121e-06,
+      "loss": 0.1159,
+      "step": 10953
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 8.553301522138992e-06,
+      "loss": 0.1163,
+      "step": 10954
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 8.553168235011863e-06,
+      "loss": 0.0702,
+      "step": 10955
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 8.553034947884734e-06,
+      "loss": 0.1246,
+      "step": 10956
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 8.552901660757604e-06,
+      "loss": 0.1386,
+      "step": 10957
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 8.552768373630476e-06,
+      "loss": 0.1222,
+      "step": 10958
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 8.552635086503346e-06,
+      "loss": 0.0908,
+      "step": 10959
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 8.552501799376217e-06,
+      "loss": 0.0766,
+      "step": 10960
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 8.552368512249088e-06,
+      "loss": 0.1046,
+      "step": 10961
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 8.552235225121959e-06,
+      "loss": 0.112,
+      "step": 10962
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 8.55210193799483e-06,
+      "loss": 0.0815,
+      "step": 10963
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 8.5519686508677e-06,
+      "loss": 0.1299,
+      "step": 10964
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 8.55183536374057e-06,
+      "loss": 0.1053,
+      "step": 10965
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 8.551702076613443e-06,
+      "loss": 0.1358,
+      "step": 10966
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 8.551568789486312e-06,
+      "loss": 0.118,
+      "step": 10967
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 8.551435502359183e-06,
+      "loss": 0.1051,
+      "step": 10968
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 8.551302215232054e-06,
+      "loss": 0.0732,
+      "step": 10969
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 8.551168928104925e-06,
+      "loss": 0.111,
+      "step": 10970
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 8.551035640977796e-06,
+      "loss": 0.1106,
+      "step": 10971
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 8.550902353850665e-06,
+      "loss": 0.0644,
+      "step": 10972
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 8.550769066723536e-06,
+      "loss": 0.1035,
+      "step": 10973
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 8.550635779596407e-06,
+      "loss": 0.1042,
+      "step": 10974
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 8.550502492469278e-06,
+      "loss": 0.1177,
+      "step": 10975
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 8.550369205342149e-06,
+      "loss": 0.0873,
+      "step": 10976
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 8.55023591821502e-06,
+      "loss": 0.0865,
+      "step": 10977
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 8.55010263108789e-06,
+      "loss": 0.0891,
+      "step": 10978
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 8.54996934396076e-06,
+      "loss": 0.1348,
+      "step": 10979
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 8.549836056833631e-06,
+      "loss": 0.1114,
+      "step": 10980
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 8.549702769706502e-06,
+      "loss": 0.0981,
+      "step": 10981
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 8.549569482579373e-06,
+      "loss": 0.0888,
+      "step": 10982
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 8.549436195452244e-06,
+      "loss": 0.1012,
+      "step": 10983
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 8.549302908325114e-06,
+      "loss": 0.1142,
+      "step": 10984
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 8.549169621197986e-06,
+      "loss": 0.0902,
+      "step": 10985
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 8.549036334070856e-06,
+      "loss": 0.1151,
+      "step": 10986
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 8.548903046943727e-06,
+      "loss": 0.1227,
+      "step": 10987
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 8.548769759816598e-06,
+      "loss": 0.0798,
+      "step": 10988
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 8.548636472689469e-06,
+      "loss": 0.0988,
+      "step": 10989
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 8.54850318556234e-06,
+      "loss": 0.0837,
+      "step": 10990
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 8.54836989843521e-06,
+      "loss": 0.0955,
+      "step": 10991
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 8.54823661130808e-06,
+      "loss": 0.1415,
+      "step": 10992
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 8.548103324180951e-06,
+      "loss": 0.0845,
+      "step": 10993
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 8.547970037053822e-06,
+      "loss": 0.1017,
+      "step": 10994
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 8.547836749926693e-06,
+      "loss": 0.0836,
+      "step": 10995
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 8.547703462799564e-06,
+      "loss": 0.1139,
+      "step": 10996
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 8.547570175672433e-06,
+      "loss": 0.1143,
+      "step": 10997
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 8.547436888545306e-06,
+      "loss": 0.0733,
+      "step": 10998
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 8.547303601418175e-06,
+      "loss": 0.1415,
+      "step": 10999
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 8.547170314291046e-06,
+      "loss": 0.0937,
+      "step": 11000
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 8.547037027163917e-06,
+      "loss": 0.1011,
+      "step": 11001
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 8.546903740036788e-06,
+      "loss": 0.0857,
+      "step": 11002
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 8.54677045290966e-06,
+      "loss": 0.0836,
+      "step": 11003
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 8.54663716578253e-06,
+      "loss": 0.0998,
+      "step": 11004
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 8.5465038786554e-06,
+      "loss": 0.0958,
+      "step": 11005
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 8.54637059152827e-06,
+      "loss": 0.1149,
+      "step": 11006
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 8.546237304401141e-06,
+      "loss": 0.1042,
+      "step": 11007
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 8.546104017274012e-06,
+      "loss": 0.0832,
+      "step": 11008
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 8.545970730146883e-06,
+      "loss": 0.0745,
+      "step": 11009
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 8.545837443019754e-06,
+      "loss": 0.1133,
+      "step": 11010
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 8.545704155892624e-06,
+      "loss": 0.0714,
+      "step": 11011
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 8.545570868765495e-06,
+      "loss": 0.0612,
+      "step": 11012
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 8.545437581638366e-06,
+      "loss": 0.1189,
+      "step": 11013
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 8.545304294511237e-06,
+      "loss": 0.117,
+      "step": 11014
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 8.545171007384108e-06,
+      "loss": 0.1149,
+      "step": 11015
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 8.545037720256977e-06,
+      "loss": 0.1021,
+      "step": 11016
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 8.54490443312985e-06,
+      "loss": 0.1139,
+      "step": 11017
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 8.544771146002719e-06,
+      "loss": 0.0714,
+      "step": 11018
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 8.54463785887559e-06,
+      "loss": 0.0764,
+      "step": 11019
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 8.544504571748461e-06,
+      "loss": 0.0638,
+      "step": 11020
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 8.544371284621332e-06,
+      "loss": 0.1697,
+      "step": 11021
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 8.544237997494203e-06,
+      "loss": 0.0875,
+      "step": 11022
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 8.544104710367074e-06,
+      "loss": 0.0874,
+      "step": 11023
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 8.543971423239943e-06,
+      "loss": 0.1383,
+      "step": 11024
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 8.543838136112816e-06,
+      "loss": 0.0853,
+      "step": 11025
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 8.543704848985685e-06,
+      "loss": 0.0966,
+      "step": 11026
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 8.543571561858556e-06,
+      "loss": 0.1068,
+      "step": 11027
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 8.543438274731427e-06,
+      "loss": 0.1029,
+      "step": 11028
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 8.543304987604298e-06,
+      "loss": 0.0809,
+      "step": 11029
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 8.54317170047717e-06,
+      "loss": 0.1463,
+      "step": 11030
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 8.543038413350039e-06,
+      "loss": 0.0901,
+      "step": 11031
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 8.54290512622291e-06,
+      "loss": 0.084,
+      "step": 11032
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 8.54277183909578e-06,
+      "loss": 0.0916,
+      "step": 11033
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 8.542638551968652e-06,
+      "loss": 0.1019,
+      "step": 11034
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 8.542505264841523e-06,
+      "loss": 0.0885,
+      "step": 11035
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 8.542371977714394e-06,
+      "loss": 0.1261,
+      "step": 11036
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 8.542238690587263e-06,
+      "loss": 0.0766,
+      "step": 11037
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 8.542105403460136e-06,
+      "loss": 0.0806,
+      "step": 11038
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 8.541972116333005e-06,
+      "loss": 0.1232,
+      "step": 11039
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 8.541838829205876e-06,
+      "loss": 0.0983,
+      "step": 11040
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 8.541705542078747e-06,
+      "loss": 0.1477,
+      "step": 11041
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 8.541572254951618e-06,
+      "loss": 0.0911,
+      "step": 11042
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 8.541438967824489e-06,
+      "loss": 0.1161,
+      "step": 11043
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 8.54130568069736e-06,
+      "loss": 0.129,
+      "step": 11044
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 8.541172393570229e-06,
+      "loss": 0.114,
+      "step": 11045
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 8.5410391064431e-06,
+      "loss": 0.1046,
+      "step": 11046
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 8.540905819315971e-06,
+      "loss": 0.0797,
+      "step": 11047
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 8.540772532188842e-06,
+      "loss": 0.1038,
+      "step": 11048
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 8.540639245061713e-06,
+      "loss": 0.1328,
+      "step": 11049
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 8.540505957934584e-06,
+      "loss": 0.0994,
+      "step": 11050
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 8.540372670807453e-06,
+      "loss": 0.0967,
+      "step": 11051
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 8.540239383680324e-06,
+      "loss": 0.0888,
+      "step": 11052
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 8.540106096553195e-06,
+      "loss": 0.1271,
+      "step": 11053
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 8.539972809426066e-06,
+      "loss": 0.1022,
+      "step": 11054
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 8.539839522298937e-06,
+      "loss": 0.1268,
+      "step": 11055
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 8.539706235171807e-06,
+      "loss": 0.1157,
+      "step": 11056
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 8.53957294804468e-06,
+      "loss": 0.1057,
+      "step": 11057
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 8.539439660917549e-06,
+      "loss": 0.1359,
+      "step": 11058
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 8.53930637379042e-06,
+      "loss": 0.1289,
+      "step": 11059
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 8.53917308666329e-06,
+      "loss": 0.1059,
+      "step": 11060
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 8.539039799536162e-06,
+      "loss": 0.1211,
+      "step": 11061
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 8.538906512409033e-06,
+      "loss": 0.1087,
+      "step": 11062
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 8.538773225281904e-06,
+      "loss": 0.1164,
+      "step": 11063
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 8.538639938154773e-06,
+      "loss": 0.0808,
+      "step": 11064
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 8.538506651027646e-06,
+      "loss": 0.1098,
+      "step": 11065
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 8.538373363900515e-06,
+      "loss": 0.0675,
+      "step": 11066
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 8.538240076773386e-06,
+      "loss": 0.1146,
+      "step": 11067
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 8.538106789646257e-06,
+      "loss": 0.0876,
+      "step": 11068
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 8.537973502519128e-06,
+      "loss": 0.0739,
+      "step": 11069
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 8.537840215391999e-06,
+      "loss": 0.0984,
+      "step": 11070
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 8.537706928264868e-06,
+      "loss": 0.0994,
+      "step": 11071
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 8.53757364113774e-06,
+      "loss": 0.1231,
+      "step": 11072
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 8.53744035401061e-06,
+      "loss": 0.1286,
+      "step": 11073
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 8.537307066883481e-06,
+      "loss": 0.0824,
+      "step": 11074
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 8.537173779756352e-06,
+      "loss": 0.0843,
+      "step": 11075
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 8.537040492629223e-06,
+      "loss": 0.0774,
+      "step": 11076
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 8.536907205502092e-06,
+      "loss": 0.0849,
+      "step": 11077
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 8.536773918374963e-06,
+      "loss": 0.0885,
+      "step": 11078
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 8.536640631247834e-06,
+      "loss": 0.0819,
+      "step": 11079
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 8.536507344120705e-06,
+      "loss": 0.1407,
+      "step": 11080
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 8.536374056993576e-06,
+      "loss": 0.1097,
+      "step": 11081
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 8.536240769866447e-06,
+      "loss": 0.0824,
+      "step": 11082
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 8.536107482739317e-06,
+      "loss": 0.1361,
+      "step": 11083
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 8.53597419561219e-06,
+      "loss": 0.1176,
+      "step": 11084
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 8.535840908485059e-06,
+      "loss": 0.086,
+      "step": 11085
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 8.53570762135793e-06,
+      "loss": 0.0802,
+      "step": 11086
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 8.5355743342308e-06,
+      "loss": 0.1288,
+      "step": 11087
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 8.535441047103672e-06,
+      "loss": 0.0988,
+      "step": 11088
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 8.535307759976543e-06,
+      "loss": 0.1452,
+      "step": 11089
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 8.535174472849412e-06,
+      "loss": 0.0518,
+      "step": 11090
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 8.535041185722283e-06,
+      "loss": 0.1282,
+      "step": 11091
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 8.534907898595154e-06,
+      "loss": 0.0921,
+      "step": 11092
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 8.534774611468025e-06,
+      "loss": 0.0807,
+      "step": 11093
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 8.534641324340896e-06,
+      "loss": 0.088,
+      "step": 11094
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 8.534508037213767e-06,
+      "loss": 0.09,
+      "step": 11095
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 8.534374750086636e-06,
+      "loss": 0.0609,
+      "step": 11096
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 8.534241462959509e-06,
+      "loss": 0.0707,
+      "step": 11097
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 8.534108175832378e-06,
+      "loss": 0.0922,
+      "step": 11098
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 8.53397488870525e-06,
+      "loss": 0.0909,
+      "step": 11099
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 8.53384160157812e-06,
+      "loss": 0.0729,
+      "step": 11100
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 8.533708314450991e-06,
+      "loss": 0.1229,
+      "step": 11101
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 8.533575027323862e-06,
+      "loss": 0.0719,
+      "step": 11102
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 8.533441740196733e-06,
+      "loss": 0.0748,
+      "step": 11103
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 8.533308453069603e-06,
+      "loss": 0.0919,
+      "step": 11104
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 8.533175165942475e-06,
+      "loss": 0.0965,
+      "step": 11105
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 8.533041878815345e-06,
+      "loss": 0.1223,
+      "step": 11106
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 8.532908591688216e-06,
+      "loss": 0.0851,
+      "step": 11107
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 8.532775304561087e-06,
+      "loss": 0.126,
+      "step": 11108
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 8.532642017433958e-06,
+      "loss": 0.0822,
+      "step": 11109
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 8.532508730306829e-06,
+      "loss": 0.075,
+      "step": 11110
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 8.532375443179698e-06,
+      "loss": 0.102,
+      "step": 11111
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 8.532242156052569e-06,
+      "loss": 0.0846,
+      "step": 11112
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 8.53210886892544e-06,
+      "loss": 0.1143,
+      "step": 11113
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 8.53197558179831e-06,
+      "loss": 0.1398,
+      "step": 11114
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 8.531842294671182e-06,
+      "loss": 0.1072,
+      "step": 11115
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 8.531709007544053e-06,
+      "loss": 0.1013,
+      "step": 11116
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 8.531575720416922e-06,
+      "loss": 0.091,
+      "step": 11117
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 8.531442433289793e-06,
+      "loss": 0.0882,
+      "step": 11118
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 8.531309146162664e-06,
+      "loss": 0.0754,
+      "step": 11119
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 8.531175859035535e-06,
+      "loss": 0.1149,
+      "step": 11120
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 8.531042571908406e-06,
+      "loss": 0.0835,
+      "step": 11121
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 8.530909284781277e-06,
+      "loss": 0.069,
+      "step": 11122
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 8.530775997654146e-06,
+      "loss": 0.0745,
+      "step": 11123
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 8.530642710527019e-06,
+      "loss": 0.0895,
+      "step": 11124
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 8.530509423399888e-06,
+      "loss": 0.0821,
+      "step": 11125
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 8.53037613627276e-06,
+      "loss": 0.0958,
+      "step": 11126
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 8.53024284914563e-06,
+      "loss": 0.1055,
+      "step": 11127
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 8.530109562018501e-06,
+      "loss": 0.1009,
+      "step": 11128
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 8.529976274891372e-06,
+      "loss": 0.0969,
+      "step": 11129
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 8.529842987764242e-06,
+      "loss": 0.0612,
+      "step": 11130
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 8.529709700637113e-06,
+      "loss": 0.0868,
+      "step": 11131
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 8.529576413509984e-06,
+      "loss": 0.0932,
+      "step": 11132
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 8.529443126382855e-06,
+      "loss": 0.1115,
+      "step": 11133
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 8.529309839255726e-06,
+      "loss": 0.0982,
+      "step": 11134
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 8.529176552128597e-06,
+      "loss": 0.1233,
+      "step": 11135
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 8.529043265001466e-06,
+      "loss": 0.105,
+      "step": 11136
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 8.528909977874339e-06,
+      "loss": 0.1119,
+      "step": 11137
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 8.528776690747208e-06,
+      "loss": 0.0968,
+      "step": 11138
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 8.528643403620079e-06,
+      "loss": 0.09,
+      "step": 11139
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 8.52851011649295e-06,
+      "loss": 0.1261,
+      "step": 11140
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 8.528376829365821e-06,
+      "loss": 0.096,
+      "step": 11141
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 8.528243542238692e-06,
+      "loss": 0.1201,
+      "step": 11142
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 8.528110255111563e-06,
+      "loss": 0.0771,
+      "step": 11143
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 8.527976967984432e-06,
+      "loss": 0.0742,
+      "step": 11144
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 8.527843680857303e-06,
+      "loss": 0.1685,
+      "step": 11145
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 8.527710393730174e-06,
+      "loss": 0.0834,
+      "step": 11146
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 8.527577106603045e-06,
+      "loss": 0.0806,
+      "step": 11147
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 8.527443819475916e-06,
+      "loss": 0.1156,
+      "step": 11148
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 8.527310532348785e-06,
+      "loss": 0.111,
+      "step": 11149
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 8.527177245221656e-06,
+      "loss": 0.1004,
+      "step": 11150
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 8.527043958094527e-06,
+      "loss": 0.1044,
+      "step": 11151
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 8.526910670967398e-06,
+      "loss": 0.0866,
+      "step": 11152
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 8.52677738384027e-06,
+      "loss": 0.0952,
+      "step": 11153
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 8.52664409671314e-06,
+      "loss": 0.1071,
+      "step": 11154
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 8.52651080958601e-06,
+      "loss": 0.0973,
+      "step": 11155
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 8.526377522458882e-06,
+      "loss": 0.1126,
+      "step": 11156
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 8.526244235331752e-06,
+      "loss": 0.0803,
+      "step": 11157
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 8.526110948204623e-06,
+      "loss": 0.1144,
+      "step": 11158
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 8.525977661077494e-06,
+      "loss": 0.1355,
+      "step": 11159
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 8.525844373950365e-06,
+      "loss": 0.0991,
+      "step": 11160
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 8.525711086823236e-06,
+      "loss": 0.1235,
+      "step": 11161
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 8.525577799696107e-06,
+      "loss": 0.0974,
+      "step": 11162
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 8.525444512568976e-06,
+      "loss": 0.1013,
+      "step": 11163
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 8.525311225441849e-06,
+      "loss": 0.0833,
+      "step": 11164
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 8.525177938314718e-06,
+      "loss": 0.101,
+      "step": 11165
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 8.525044651187589e-06,
+      "loss": 0.1258,
+      "step": 11166
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 8.52491136406046e-06,
+      "loss": 0.0801,
+      "step": 11167
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 8.52477807693333e-06,
+      "loss": 0.0828,
+      "step": 11168
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 8.524644789806202e-06,
+      "loss": 0.1086,
+      "step": 11169
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 8.524511502679071e-06,
+      "loss": 0.0878,
+      "step": 11170
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 8.524378215551942e-06,
+      "loss": 0.0909,
+      "step": 11171
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 8.524244928424813e-06,
+      "loss": 0.0772,
+      "step": 11172
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 8.524111641297684e-06,
+      "loss": 0.1006,
+      "step": 11173
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 8.523978354170555e-06,
+      "loss": 0.0802,
+      "step": 11174
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 8.523845067043426e-06,
+      "loss": 0.1408,
+      "step": 11175
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 8.523711779916296e-06,
+      "loss": 0.0827,
+      "step": 11176
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 8.523578492789167e-06,
+      "loss": 0.0749,
+      "step": 11177
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 8.523445205662038e-06,
+      "loss": 0.0896,
+      "step": 11178
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 8.523311918534909e-06,
+      "loss": 0.0894,
+      "step": 11179
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 8.52317863140778e-06,
+      "loss": 0.1184,
+      "step": 11180
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 8.52304534428065e-06,
+      "loss": 0.1187,
+      "step": 11181
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 8.52291205715352e-06,
+      "loss": 0.1181,
+      "step": 11182
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 8.522778770026393e-06,
+      "loss": 0.0961,
+      "step": 11183
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 8.522645482899262e-06,
+      "loss": 0.1437,
+      "step": 11184
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 8.522512195772133e-06,
+      "loss": 0.1307,
+      "step": 11185
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 8.522378908645004e-06,
+      "loss": 0.0837,
+      "step": 11186
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 8.522245621517875e-06,
+      "loss": 0.111,
+      "step": 11187
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 8.522112334390746e-06,
+      "loss": 0.1434,
+      "step": 11188
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 8.521979047263615e-06,
+      "loss": 0.1449,
+      "step": 11189
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 8.521845760136486e-06,
+      "loss": 0.0985,
+      "step": 11190
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 8.521712473009357e-06,
+      "loss": 0.1354,
+      "step": 11191
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 8.521579185882228e-06,
+      "loss": 0.0733,
+      "step": 11192
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 8.521445898755099e-06,
+      "loss": 0.1005,
+      "step": 11193
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 8.52131261162797e-06,
+      "loss": 0.0886,
+      "step": 11194
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 8.52117932450084e-06,
+      "loss": 0.0991,
+      "step": 11195
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 8.521046037373712e-06,
+      "loss": 0.118,
+      "step": 11196
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 8.520912750246581e-06,
+      "loss": 0.0726,
+      "step": 11197
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 8.520779463119452e-06,
+      "loss": 0.133,
+      "step": 11198
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 8.520646175992323e-06,
+      "loss": 0.0824,
+      "step": 11199
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 8.520512888865194e-06,
+      "loss": 0.0771,
+      "step": 11200
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 8.520379601738065e-06,
+      "loss": 0.099,
+      "step": 11201
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 8.520246314610936e-06,
+      "loss": 0.1117,
+      "step": 11202
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 8.520113027483806e-06,
+      "loss": 0.1027,
+      "step": 11203
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 8.519979740356678e-06,
+      "loss": 0.1206,
+      "step": 11204
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 8.519846453229548e-06,
+      "loss": 0.0894,
+      "step": 11205
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 8.519713166102419e-06,
+      "loss": 0.1178,
+      "step": 11206
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 8.51957987897529e-06,
+      "loss": 0.1182,
+      "step": 11207
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 8.519446591848159e-06,
+      "loss": 0.0901,
+      "step": 11208
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 8.519313304721032e-06,
+      "loss": 0.1078,
+      "step": 11209
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 8.519180017593901e-06,
+      "loss": 0.0972,
+      "step": 11210
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 8.519046730466772e-06,
+      "loss": 0.0933,
+      "step": 11211
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 8.518913443339643e-06,
+      "loss": 0.0871,
+      "step": 11212
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 8.518780156212514e-06,
+      "loss": 0.1008,
+      "step": 11213
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 8.518646869085385e-06,
+      "loss": 0.0926,
+      "step": 11214
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 8.518513581958256e-06,
+      "loss": 0.1305,
+      "step": 11215
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 8.518380294831125e-06,
+      "loss": 0.0848,
+      "step": 11216
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 8.518247007703996e-06,
+      "loss": 0.1079,
+      "step": 11217
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 8.518113720576867e-06,
+      "loss": 0.078,
+      "step": 11218
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 8.517980433449738e-06,
+      "loss": 0.0755,
+      "step": 11219
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 8.517847146322609e-06,
+      "loss": 0.1074,
+      "step": 11220
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 8.51771385919548e-06,
+      "loss": 0.0877,
+      "step": 11221
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 8.51758057206835e-06,
+      "loss": 0.1199,
+      "step": 11222
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 8.517447284941222e-06,
+      "loss": 0.0763,
+      "step": 11223
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 8.517313997814091e-06,
+      "loss": 0.0764,
+      "step": 11224
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 8.517180710686962e-06,
+      "loss": 0.0917,
+      "step": 11225
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 8.517047423559833e-06,
+      "loss": 0.1535,
+      "step": 11226
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 8.516914136432703e-06,
+      "loss": 0.0914,
+      "step": 11227
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 8.516780849305575e-06,
+      "loss": 0.0926,
+      "step": 11228
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 8.516647562178445e-06,
+      "loss": 0.1501,
+      "step": 11229
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 8.516514275051316e-06,
+      "loss": 0.0835,
+      "step": 11230
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 8.516380987924187e-06,
+      "loss": 0.0853,
+      "step": 11231
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 8.516247700797058e-06,
+      "loss": 0.1145,
+      "step": 11232
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 8.516114413669929e-06,
+      "loss": 0.0988,
+      "step": 11233
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 8.5159811265428e-06,
+      "loss": 0.1413,
+      "step": 11234
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 8.515847839415669e-06,
+      "loss": 0.1137,
+      "step": 11235
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 8.515714552288542e-06,
+      "loss": 0.1017,
+      "step": 11236
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 8.515581265161411e-06,
+      "loss": 0.1258,
+      "step": 11237
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 8.515447978034282e-06,
+      "loss": 0.1038,
+      "step": 11238
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 8.515314690907153e-06,
+      "loss": 0.1419,
+      "step": 11239
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 8.515181403780024e-06,
+      "loss": 0.1025,
+      "step": 11240
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 8.515048116652895e-06,
+      "loss": 0.1198,
+      "step": 11241
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 8.514914829525766e-06,
+      "loss": 0.0965,
+      "step": 11242
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 8.514781542398635e-06,
+      "loss": 0.1059,
+      "step": 11243
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 8.514648255271506e-06,
+      "loss": 0.1118,
+      "step": 11244
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 8.514514968144377e-06,
+      "loss": 0.1546,
+      "step": 11245
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 8.514381681017248e-06,
+      "loss": 0.0907,
+      "step": 11246
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 8.51424839389012e-06,
+      "loss": 0.1211,
+      "step": 11247
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 8.514115106762989e-06,
+      "loss": 0.1186,
+      "step": 11248
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 8.51398181963586e-06,
+      "loss": 0.1059,
+      "step": 11249
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 8.51384853250873e-06,
+      "loss": 0.0937,
+      "step": 11250
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 8.513715245381602e-06,
+      "loss": 0.0708,
+      "step": 11251
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 8.513581958254473e-06,
+      "loss": 0.1179,
+      "step": 11252
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 8.513448671127344e-06,
+      "loss": 0.1288,
+      "step": 11253
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 8.513315384000213e-06,
+      "loss": 0.1165,
+      "step": 11254
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 8.513182096873085e-06,
+      "loss": 0.1199,
+      "step": 11255
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 8.513048809745955e-06,
+      "loss": 0.0861,
+      "step": 11256
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 8.512915522618826e-06,
+      "loss": 0.0802,
+      "step": 11257
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 8.512782235491697e-06,
+      "loss": 0.1064,
+      "step": 11258
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 8.512648948364568e-06,
+      "loss": 0.0985,
+      "step": 11259
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 8.512515661237439e-06,
+      "loss": 0.0767,
+      "step": 11260
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 8.51238237411031e-06,
+      "loss": 0.0882,
+      "step": 11261
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 8.512249086983179e-06,
+      "loss": 0.1061,
+      "step": 11262
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 8.512115799856052e-06,
+      "loss": 0.1153,
+      "step": 11263
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 8.511982512728921e-06,
+      "loss": 0.088,
+      "step": 11264
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 8.511849225601792e-06,
+      "loss": 0.1018,
+      "step": 11265
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 8.511715938474663e-06,
+      "loss": 0.123,
+      "step": 11266
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 8.511582651347532e-06,
+      "loss": 0.0845,
+      "step": 11267
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 8.511449364220405e-06,
+      "loss": 0.0974,
+      "step": 11268
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 8.511316077093274e-06,
+      "loss": 0.1382,
+      "step": 11269
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 8.511182789966145e-06,
+      "loss": 0.0974,
+      "step": 11270
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 8.511049502839016e-06,
+      "loss": 0.0831,
+      "step": 11271
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 8.510916215711887e-06,
+      "loss": 0.0782,
+      "step": 11272
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 8.510782928584758e-06,
+      "loss": 0.0932,
+      "step": 11273
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 8.51064964145763e-06,
+      "loss": 0.1674,
+      "step": 11274
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 8.510516354330499e-06,
+      "loss": 0.125,
+      "step": 11275
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 8.510383067203371e-06,
+      "loss": 0.0769,
+      "step": 11276
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 8.51024978007624e-06,
+      "loss": 0.0876,
+      "step": 11277
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 8.510116492949112e-06,
+      "loss": 0.1294,
+      "step": 11278
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 8.509983205821983e-06,
+      "loss": 0.0723,
+      "step": 11279
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 8.509849918694854e-06,
+      "loss": 0.1016,
+      "step": 11280
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 8.509716631567725e-06,
+      "loss": 0.083,
+      "step": 11281
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 8.509583344440596e-06,
+      "loss": 0.1083,
+      "step": 11282
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 8.509450057313465e-06,
+      "loss": 0.0958,
+      "step": 11283
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 8.509316770186336e-06,
+      "loss": 0.0973,
+      "step": 11284
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 8.509183483059207e-06,
+      "loss": 0.0924,
+      "step": 11285
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 8.509050195932078e-06,
+      "loss": 0.1173,
+      "step": 11286
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 8.508916908804949e-06,
+      "loss": 0.09,
+      "step": 11287
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 8.508783621677818e-06,
+      "loss": 0.0901,
+      "step": 11288
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 8.508650334550689e-06,
+      "loss": 0.1115,
+      "step": 11289
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 8.50851704742356e-06,
+      "loss": 0.0706,
+      "step": 11290
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 8.508383760296431e-06,
+      "loss": 0.1264,
+      "step": 11291
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 8.508250473169302e-06,
+      "loss": 0.1181,
+      "step": 11292
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 8.508117186042173e-06,
+      "loss": 0.1062,
+      "step": 11293
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 8.507983898915042e-06,
+      "loss": 0.0896,
+      "step": 11294
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 8.507850611787915e-06,
+      "loss": 0.0886,
+      "step": 11295
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 8.507717324660784e-06,
+      "loss": 0.0947,
+      "step": 11296
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 8.507584037533655e-06,
+      "loss": 0.102,
+      "step": 11297
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 8.507450750406526e-06,
+      "loss": 0.1312,
+      "step": 11298
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 8.507317463279397e-06,
+      "loss": 0.1085,
+      "step": 11299
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 8.507184176152268e-06,
+      "loss": 0.1183,
+      "step": 11300
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 8.50705088902514e-06,
+      "loss": 0.0962,
+      "step": 11301
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 8.506917601898009e-06,
+      "loss": 0.0605,
+      "step": 11302
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 8.506784314770881e-06,
+      "loss": 0.1287,
+      "step": 11303
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 8.50665102764375e-06,
+      "loss": 0.1254,
+      "step": 11304
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 8.506517740516622e-06,
+      "loss": 0.0917,
+      "step": 11305
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 8.506384453389493e-06,
+      "loss": 0.1618,
+      "step": 11306
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 8.506251166262362e-06,
+      "loss": 0.1036,
+      "step": 11307
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 8.506117879135235e-06,
+      "loss": 0.127,
+      "step": 11308
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 8.505984592008104e-06,
+      "loss": 0.1087,
+      "step": 11309
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 8.505851304880975e-06,
+      "loss": 0.1309,
+      "step": 11310
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 8.505718017753846e-06,
+      "loss": 0.1286,
+      "step": 11311
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 8.505584730626717e-06,
+      "loss": 0.0915,
+      "step": 11312
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 8.505451443499588e-06,
+      "loss": 0.142,
+      "step": 11313
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 8.505318156372459e-06,
+      "loss": 0.1108,
+      "step": 11314
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 8.505184869245328e-06,
+      "loss": 0.1277,
+      "step": 11315
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 8.5050515821182e-06,
+      "loss": 0.1359,
+      "step": 11316
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 8.50491829499107e-06,
+      "loss": 0.1036,
+      "step": 11317
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 8.504785007863941e-06,
+      "loss": 0.1436,
+      "step": 11318
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 8.504651720736812e-06,
+      "loss": 0.0931,
+      "step": 11319
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 8.504518433609683e-06,
+      "loss": 0.0944,
+      "step": 11320
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 8.504385146482553e-06,
+      "loss": 0.1172,
+      "step": 11321
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 8.504251859355425e-06,
+      "loss": 0.1049,
+      "step": 11322
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 8.504118572228294e-06,
+      "loss": 0.0702,
+      "step": 11323
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 8.503985285101165e-06,
+      "loss": 0.1088,
+      "step": 11324
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 8.503851997974036e-06,
+      "loss": 0.1251,
+      "step": 11325
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 8.503718710846906e-06,
+      "loss": 0.0698,
+      "step": 11326
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 8.503585423719778e-06,
+      "loss": 0.0892,
+      "step": 11327
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 8.503452136592648e-06,
+      "loss": 0.0898,
+      "step": 11328
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 8.503318849465519e-06,
+      "loss": 0.1115,
+      "step": 11329
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 8.50318556233839e-06,
+      "loss": 0.1246,
+      "step": 11330
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 8.50305227521126e-06,
+      "loss": 0.0783,
+      "step": 11331
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 8.502918988084132e-06,
+      "loss": 0.0904,
+      "step": 11332
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 8.502785700957003e-06,
+      "loss": 0.109,
+      "step": 11333
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 8.502652413829872e-06,
+      "loss": 0.0828,
+      "step": 11334
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 8.502519126702745e-06,
+      "loss": 0.0918,
+      "step": 11335
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 8.502385839575614e-06,
+      "loss": 0.0991,
+      "step": 11336
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 8.502252552448485e-06,
+      "loss": 0.1185,
+      "step": 11337
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 8.502119265321356e-06,
+      "loss": 0.0807,
+      "step": 11338
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 8.501985978194227e-06,
+      "loss": 0.0792,
+      "step": 11339
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 8.501852691067098e-06,
+      "loss": 0.1136,
+      "step": 11340
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 8.501719403939969e-06,
+      "loss": 0.0914,
+      "step": 11341
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 8.501586116812838e-06,
+      "loss": 0.1318,
+      "step": 11342
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 8.501452829685711e-06,
+      "loss": 0.1005,
+      "step": 11343
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 8.50131954255858e-06,
+      "loss": 0.0837,
+      "step": 11344
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 8.501186255431451e-06,
+      "loss": 0.1135,
+      "step": 11345
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 8.501052968304322e-06,
+      "loss": 0.1203,
+      "step": 11346
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 8.500919681177192e-06,
+      "loss": 0.1136,
+      "step": 11347
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 8.500786394050064e-06,
+      "loss": 0.0912,
+      "step": 11348
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 8.500653106922934e-06,
+      "loss": 0.1312,
+      "step": 11349
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 8.500519819795805e-06,
+      "loss": 0.0874,
+      "step": 11350
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 8.500386532668676e-06,
+      "loss": 0.1248,
+      "step": 11351
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 8.500253245541547e-06,
+      "loss": 0.0937,
+      "step": 11352
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 8.500119958414418e-06,
+      "loss": 0.1149,
+      "step": 11353
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 8.499986671287289e-06,
+      "loss": 0.1388,
+      "step": 11354
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 8.499853384160158e-06,
+      "loss": 0.1337,
+      "step": 11355
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 8.499720097033029e-06,
+      "loss": 0.1021,
+      "step": 11356
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 8.4995868099059e-06,
+      "loss": 0.0959,
+      "step": 11357
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 8.49945352277877e-06,
+      "loss": 0.1515,
+      "step": 11358
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 8.499320235651642e-06,
+      "loss": 0.0754,
+      "step": 11359
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 8.499186948524513e-06,
+      "loss": 0.1174,
+      "step": 11360
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 8.499053661397382e-06,
+      "loss": 0.1237,
+      "step": 11361
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 8.498920374270255e-06,
+      "loss": 0.1218,
+      "step": 11362
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 8.498787087143124e-06,
+      "loss": 0.1007,
+      "step": 11363
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 8.498653800015995e-06,
+      "loss": 0.0844,
+      "step": 11364
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 8.498520512888866e-06,
+      "loss": 0.0926,
+      "step": 11365
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 8.498387225761735e-06,
+      "loss": 0.1084,
+      "step": 11366
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 8.498253938634608e-06,
+      "loss": 0.131,
+      "step": 11367
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 8.498120651507477e-06,
+      "loss": 0.0903,
+      "step": 11368
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 8.497987364380348e-06,
+      "loss": 0.0879,
+      "step": 11369
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 8.49785407725322e-06,
+      "loss": 0.0947,
+      "step": 11370
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 8.49772079012609e-06,
+      "loss": 0.0914,
+      "step": 11371
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 8.497587502998961e-06,
+      "loss": 0.0852,
+      "step": 11372
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 8.497454215871832e-06,
+      "loss": 0.0964,
+      "step": 11373
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 8.497320928744702e-06,
+      "loss": 0.0748,
+      "step": 11374
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 8.497187641617574e-06,
+      "loss": 0.1083,
+      "step": 11375
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 8.497054354490444e-06,
+      "loss": 0.1279,
+      "step": 11376
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 8.496921067363315e-06,
+      "loss": 0.1505,
+      "step": 11377
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 8.496787780236186e-06,
+      "loss": 0.0711,
+      "step": 11378
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 8.496654493109057e-06,
+      "loss": 0.0755,
+      "step": 11379
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 8.496521205981928e-06,
+      "loss": 0.1113,
+      "step": 11380
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 8.496387918854799e-06,
+      "loss": 0.1417,
+      "step": 11381
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 8.496254631727668e-06,
+      "loss": 0.0877,
+      "step": 11382
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 8.496121344600539e-06,
+      "loss": 0.0906,
+      "step": 11383
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 8.49598805747341e-06,
+      "loss": 0.0639,
+      "step": 11384
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 8.495854770346281e-06,
+      "loss": 0.1505,
+      "step": 11385
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 8.495721483219152e-06,
+      "loss": 0.1184,
+      "step": 11386
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 8.495588196092021e-06,
+      "loss": 0.1226,
+      "step": 11387
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 8.495454908964892e-06,
+      "loss": 0.1236,
+      "step": 11388
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 8.495321621837763e-06,
+      "loss": 0.1165,
+      "step": 11389
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 8.495188334710634e-06,
+      "loss": 0.0692,
+      "step": 11390
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 8.495055047583505e-06,
+      "loss": 0.0827,
+      "step": 11391
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 8.494921760456376e-06,
+      "loss": 0.0726,
+      "step": 11392
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 8.494788473329245e-06,
+      "loss": 0.1105,
+      "step": 11393
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 8.494655186202118e-06,
+      "loss": 0.109,
+      "step": 11394
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 8.494521899074987e-06,
+      "loss": 0.0838,
+      "step": 11395
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 8.494388611947858e-06,
+      "loss": 0.0938,
+      "step": 11396
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 8.49425532482073e-06,
+      "loss": 0.0832,
+      "step": 11397
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 8.4941220376936e-06,
+      "loss": 0.0861,
+      "step": 11398
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 8.493988750566471e-06,
+      "loss": 0.1178,
+      "step": 11399
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 8.493855463439342e-06,
+      "loss": 0.1249,
+      "step": 11400
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 8.493722176312212e-06,
+      "loss": 0.1043,
+      "step": 11401
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 8.493588889185084e-06,
+      "loss": 0.1118,
+      "step": 11402
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 8.493455602057954e-06,
+      "loss": 0.07,
+      "step": 11403
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 8.493322314930825e-06,
+      "loss": 0.0977,
+      "step": 11404
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 8.493189027803696e-06,
+      "loss": 0.1161,
+      "step": 11405
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 8.493055740676565e-06,
+      "loss": 0.108,
+      "step": 11406
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 8.492922453549438e-06,
+      "loss": 0.0991,
+      "step": 11407
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 8.492789166422307e-06,
+      "loss": 0.1517,
+      "step": 11408
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 8.492655879295178e-06,
+      "loss": 0.12,
+      "step": 11409
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 8.492522592168049e-06,
+      "loss": 0.1131,
+      "step": 11410
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 8.49238930504092e-06,
+      "loss": 0.0986,
+      "step": 11411
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 8.492256017913791e-06,
+      "loss": 0.0871,
+      "step": 11412
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 8.492122730786662e-06,
+      "loss": 0.0929,
+      "step": 11413
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 8.491989443659531e-06,
+      "loss": 0.1468,
+      "step": 11414
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 8.491856156532402e-06,
+      "loss": 0.0974,
+      "step": 11415
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 8.491722869405273e-06,
+      "loss": 0.0961,
+      "step": 11416
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 8.491589582278144e-06,
+      "loss": 0.1429,
+      "step": 11417
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 8.491456295151015e-06,
+      "loss": 0.0993,
+      "step": 11418
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 8.491323008023886e-06,
+      "loss": 0.0965,
+      "step": 11419
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 8.491189720896756e-06,
+      "loss": 0.107,
+      "step": 11420
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 8.491056433769628e-06,
+      "loss": 0.0818,
+      "step": 11421
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 8.490923146642498e-06,
+      "loss": 0.114,
+      "step": 11422
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 8.490789859515369e-06,
+      "loss": 0.1288,
+      "step": 11423
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 8.49065657238824e-06,
+      "loss": 0.0904,
+      "step": 11424
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 8.490523285261109e-06,
+      "loss": 0.1099,
+      "step": 11425
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 8.490389998133982e-06,
+      "loss": 0.0793,
+      "step": 11426
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 8.49025671100685e-06,
+      "loss": 0.0838,
+      "step": 11427
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 8.490123423879722e-06,
+      "loss": 0.0994,
+      "step": 11428
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 8.489990136752593e-06,
+      "loss": 0.1328,
+      "step": 11429
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 8.489856849625464e-06,
+      "loss": 0.1241,
+      "step": 11430
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 8.489723562498335e-06,
+      "loss": 0.103,
+      "step": 11431
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 8.489590275371206e-06,
+      "loss": 0.1142,
+      "step": 11432
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 8.489456988244075e-06,
+      "loss": 0.1048,
+      "step": 11433
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 8.489323701116948e-06,
+      "loss": 0.114,
+      "step": 11434
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 8.489190413989817e-06,
+      "loss": 0.084,
+      "step": 11435
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 8.489057126862688e-06,
+      "loss": 0.1027,
+      "step": 11436
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 8.488923839735559e-06,
+      "loss": 0.1167,
+      "step": 11437
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 8.48879055260843e-06,
+      "loss": 0.0981,
+      "step": 11438
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 8.488657265481301e-06,
+      "loss": 0.1321,
+      "step": 11439
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 8.488523978354172e-06,
+      "loss": 0.1343,
+      "step": 11440
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 8.488390691227041e-06,
+      "loss": 0.1081,
+      "step": 11441
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 8.488257404099914e-06,
+      "loss": 0.0664,
+      "step": 11442
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 8.488124116972783e-06,
+      "loss": 0.1184,
+      "step": 11443
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 8.487990829845654e-06,
+      "loss": 0.1179,
+      "step": 11444
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 8.487857542718525e-06,
+      "loss": 0.1507,
+      "step": 11445
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 8.487724255591395e-06,
+      "loss": 0.076,
+      "step": 11446
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 8.487590968464267e-06,
+      "loss": 0.1278,
+      "step": 11447
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 8.487457681337137e-06,
+      "loss": 0.1041,
+      "step": 11448
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 8.487324394210008e-06,
+      "loss": 0.11,
+      "step": 11449
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 8.487191107082879e-06,
+      "loss": 0.0978,
+      "step": 11450
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 8.48705781995575e-06,
+      "loss": 0.1125,
+      "step": 11451
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 8.48692453282862e-06,
+      "loss": 0.1302,
+      "step": 11452
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 8.486791245701492e-06,
+      "loss": 0.0945,
+      "step": 11453
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 8.486657958574361e-06,
+      "loss": 0.1335,
+      "step": 11454
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 8.486524671447232e-06,
+      "loss": 0.1109,
+      "step": 11455
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 8.486391384320103e-06,
+      "loss": 0.1375,
+      "step": 11456
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 8.486258097192974e-06,
+      "loss": 0.0783,
+      "step": 11457
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 8.486124810065845e-06,
+      "loss": 0.0719,
+      "step": 11458
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 8.485991522938716e-06,
+      "loss": 0.1109,
+      "step": 11459
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 8.485858235811585e-06,
+      "loss": 0.0928,
+      "step": 11460
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 8.485724948684458e-06,
+      "loss": 0.0921,
+      "step": 11461
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 8.485591661557327e-06,
+      "loss": 0.0846,
+      "step": 11462
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 8.485458374430198e-06,
+      "loss": 0.0881,
+      "step": 11463
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 8.48532508730307e-06,
+      "loss": 0.0653,
+      "step": 11464
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 8.485191800175938e-06,
+      "loss": 0.0795,
+      "step": 11465
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 8.485058513048811e-06,
+      "loss": 0.1096,
+      "step": 11466
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 8.48492522592168e-06,
+      "loss": 0.126,
+      "step": 11467
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 8.484791938794551e-06,
+      "loss": 0.1131,
+      "step": 11468
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 8.484658651667422e-06,
+      "loss": 0.1124,
+      "step": 11469
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 8.484525364540293e-06,
+      "loss": 0.0856,
+      "step": 11470
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 8.484392077413164e-06,
+      "loss": 0.1142,
+      "step": 11471
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 8.484258790286035e-06,
+      "loss": 0.0971,
+      "step": 11472
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 8.484125503158905e-06,
+      "loss": 0.1367,
+      "step": 11473
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 8.483992216031777e-06,
+      "loss": 0.1049,
+      "step": 11474
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 8.483858928904647e-06,
+      "loss": 0.1039,
+      "step": 11475
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 8.483725641777518e-06,
+      "loss": 0.1269,
+      "step": 11476
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 8.483592354650389e-06,
+      "loss": 0.1607,
+      "step": 11477
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 8.48345906752326e-06,
+      "loss": 0.0762,
+      "step": 11478
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 8.48332578039613e-06,
+      "loss": 0.0879,
+      "step": 11479
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 8.483192493269002e-06,
+      "loss": 0.0974,
+      "step": 11480
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 8.483059206141871e-06,
+      "loss": 0.1598,
+      "step": 11481
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 8.482925919014742e-06,
+      "loss": 0.0836,
+      "step": 11482
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 8.482792631887613e-06,
+      "loss": 0.0899,
+      "step": 11483
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 8.482659344760484e-06,
+      "loss": 0.1205,
+      "step": 11484
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 8.482526057633355e-06,
+      "loss": 0.0684,
+      "step": 11485
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 8.482392770506224e-06,
+      "loss": 0.1114,
+      "step": 11486
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 8.482259483379095e-06,
+      "loss": 0.0895,
+      "step": 11487
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 8.482126196251966e-06,
+      "loss": 0.1294,
+      "step": 11488
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 8.481992909124837e-06,
+      "loss": 0.0805,
+      "step": 11489
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 8.481859621997708e-06,
+      "loss": 0.119,
+      "step": 11490
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 8.48172633487058e-06,
+      "loss": 0.1075,
+      "step": 11491
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 8.481593047743449e-06,
+      "loss": 0.1099,
+      "step": 11492
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 8.481459760616321e-06,
+      "loss": 0.0795,
+      "step": 11493
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 8.48132647348919e-06,
+      "loss": 0.1357,
+      "step": 11494
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 8.481193186362062e-06,
+      "loss": 0.1233,
+      "step": 11495
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 8.481059899234933e-06,
+      "loss": 0.1056,
+      "step": 11496
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 8.480926612107804e-06,
+      "loss": 0.0795,
+      "step": 11497
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 8.480793324980675e-06,
+      "loss": 0.1115,
+      "step": 11498
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 8.480660037853546e-06,
+      "loss": 0.0864,
+      "step": 11499
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 8.480526750726415e-06,
+      "loss": 0.1009,
+      "step": 11500
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 8.480393463599286e-06,
+      "loss": 0.1217,
+      "step": 11501
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 8.480260176472157e-06,
+      "loss": 0.1154,
+      "step": 11502
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 8.480126889345028e-06,
+      "loss": 0.0879,
+      "step": 11503
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 8.479993602217899e-06,
+      "loss": 0.1394,
+      "step": 11504
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 8.479860315090768e-06,
+      "loss": 0.0765,
+      "step": 11505
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 8.47972702796364e-06,
+      "loss": 0.0622,
+      "step": 11506
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 8.47959374083651e-06,
+      "loss": 0.111,
+      "step": 11507
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 8.479460453709381e-06,
+      "loss": 0.1075,
+      "step": 11508
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 8.479327166582252e-06,
+      "loss": 0.0719,
+      "step": 11509
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 8.479193879455123e-06,
+      "loss": 0.0957,
+      "step": 11510
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 8.479060592327994e-06,
+      "loss": 0.1081,
+      "step": 11511
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 8.478927305200865e-06,
+      "loss": 0.1199,
+      "step": 11512
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 8.478794018073734e-06,
+      "loss": 0.098,
+      "step": 11513
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 8.478660730946607e-06,
+      "loss": 0.0712,
+      "step": 11514
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 8.478527443819476e-06,
+      "loss": 0.1187,
+      "step": 11515
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 8.478394156692347e-06,
+      "loss": 0.1394,
+      "step": 11516
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 8.478260869565218e-06,
+      "loss": 0.0981,
+      "step": 11517
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 8.47812758243809e-06,
+      "loss": 0.1116,
+      "step": 11518
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 8.47799429531096e-06,
+      "loss": 0.098,
+      "step": 11519
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 8.477861008183831e-06,
+      "loss": 0.1098,
+      "step": 11520
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 8.4777277210567e-06,
+      "loss": 0.0998,
+      "step": 11521
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 8.477594433929572e-06,
+      "loss": 0.1137,
+      "step": 11522
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 8.477461146802443e-06,
+      "loss": 0.0901,
+      "step": 11523
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 8.477327859675314e-06,
+      "loss": 0.0903,
+      "step": 11524
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 8.477194572548185e-06,
+      "loss": 0.0873,
+      "step": 11525
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 8.477061285421054e-06,
+      "loss": 0.119,
+      "step": 11526
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 8.476927998293925e-06,
+      "loss": 0.136,
+      "step": 11527
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 8.476794711166796e-06,
+      "loss": 0.1101,
+      "step": 11528
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 8.476661424039667e-06,
+      "loss": 0.0717,
+      "step": 11529
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 8.476528136912538e-06,
+      "loss": 0.0776,
+      "step": 11530
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 8.476394849785409e-06,
+      "loss": 0.1229,
+      "step": 11531
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 8.476261562658278e-06,
+      "loss": 0.1384,
+      "step": 11532
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 8.476128275531151e-06,
+      "loss": 0.1156,
+      "step": 11533
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 8.47599498840402e-06,
+      "loss": 0.0705,
+      "step": 11534
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 8.475861701276891e-06,
+      "loss": 0.1381,
+      "step": 11535
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 8.475728414149762e-06,
+      "loss": 0.1122,
+      "step": 11536
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 8.475595127022633e-06,
+      "loss": 0.1021,
+      "step": 11537
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 8.475461839895504e-06,
+      "loss": 0.1042,
+      "step": 11538
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 8.475328552768375e-06,
+      "loss": 0.079,
+      "step": 11539
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 8.475195265641244e-06,
+      "loss": 0.0967,
+      "step": 11540
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 8.475061978514115e-06,
+      "loss": 0.1009,
+      "step": 11541
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 8.474928691386986e-06,
+      "loss": 0.1186,
+      "step": 11542
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 8.474795404259857e-06,
+      "loss": 0.0957,
+      "step": 11543
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 8.474662117132728e-06,
+      "loss": 0.0532,
+      "step": 11544
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 8.474528830005598e-06,
+      "loss": 0.0551,
+      "step": 11545
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 8.47439554287847e-06,
+      "loss": 0.109,
+      "step": 11546
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 8.47426225575134e-06,
+      "loss": 0.0962,
+      "step": 11547
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 8.47412896862421e-06,
+      "loss": 0.1142,
+      "step": 11548
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 8.473995681497082e-06,
+      "loss": 0.098,
+      "step": 11549
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 8.473862394369953e-06,
+      "loss": 0.0985,
+      "step": 11550
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 8.473729107242824e-06,
+      "loss": 0.1129,
+      "step": 11551
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 8.473595820115695e-06,
+      "loss": 0.075,
+      "step": 11552
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 8.473462532988564e-06,
+      "loss": 0.0975,
+      "step": 11553
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 8.473329245861435e-06,
+      "loss": 0.116,
+      "step": 11554
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 8.473195958734306e-06,
+      "loss": 0.1136,
+      "step": 11555
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 8.473062671607177e-06,
+      "loss": 0.1007,
+      "step": 11556
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 8.472929384480048e-06,
+      "loss": 0.0882,
+      "step": 11557
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 8.472796097352919e-06,
+      "loss": 0.1051,
+      "step": 11558
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 8.472662810225788e-06,
+      "loss": 0.0932,
+      "step": 11559
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 8.47252952309866e-06,
+      "loss": 0.0657,
+      "step": 11560
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 8.47239623597153e-06,
+      "loss": 0.1213,
+      "step": 11561
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 8.472262948844401e-06,
+      "loss": 0.1055,
+      "step": 11562
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 8.472129661717272e-06,
+      "loss": 0.1399,
+      "step": 11563
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 8.471996374590142e-06,
+      "loss": 0.0814,
+      "step": 11564
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 8.471863087463014e-06,
+      "loss": 0.1033,
+      "step": 11565
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 8.471729800335884e-06,
+      "loss": 0.075,
+      "step": 11566
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 8.471596513208755e-06,
+      "loss": 0.1027,
+      "step": 11567
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 8.471463226081626e-06,
+      "loss": 0.0987,
+      "step": 11568
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 8.471329938954497e-06,
+      "loss": 0.1278,
+      "step": 11569
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 8.471196651827368e-06,
+      "loss": 0.1313,
+      "step": 11570
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 8.471063364700239e-06,
+      "loss": 0.1254,
+      "step": 11571
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 8.470930077573108e-06,
+      "loss": 0.0853,
+      "step": 11572
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 8.47079679044598e-06,
+      "loss": 0.1377,
+      "step": 11573
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 8.47066350331885e-06,
+      "loss": 0.1092,
+      "step": 11574
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 8.47053021619172e-06,
+      "loss": 0.1396,
+      "step": 11575
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 8.470396929064592e-06,
+      "loss": 0.1052,
+      "step": 11576
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 8.470263641937463e-06,
+      "loss": 0.104,
+      "step": 11577
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 8.470130354810334e-06,
+      "loss": 0.1099,
+      "step": 11578
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 8.469997067683205e-06,
+      "loss": 0.0922,
+      "step": 11579
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 8.469863780556074e-06,
+      "loss": 0.1366,
+      "step": 11580
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 8.469730493428945e-06,
+      "loss": 0.1137,
+      "step": 11581
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 8.469597206301816e-06,
+      "loss": 0.1052,
+      "step": 11582
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 8.469463919174687e-06,
+      "loss": 0.064,
+      "step": 11583
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 8.469330632047558e-06,
+      "loss": 0.0796,
+      "step": 11584
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 8.469197344920427e-06,
+      "loss": 0.1403,
+      "step": 11585
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 8.4690640577933e-06,
+      "loss": 0.0951,
+      "step": 11586
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 8.46893077066617e-06,
+      "loss": 0.0943,
+      "step": 11587
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 8.46879748353904e-06,
+      "loss": 0.1044,
+      "step": 11588
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 8.468664196411911e-06,
+      "loss": 0.105,
+      "step": 11589
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 8.468530909284782e-06,
+      "loss": 0.1384,
+      "step": 11590
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 8.468397622157653e-06,
+      "loss": 0.1327,
+      "step": 11591
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 8.468264335030524e-06,
+      "loss": 0.1139,
+      "step": 11592
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 8.468131047903394e-06,
+      "loss": 0.0948,
+      "step": 11593
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 8.467997760776265e-06,
+      "loss": 0.0803,
+      "step": 11594
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 8.467864473649136e-06,
+      "loss": 0.0883,
+      "step": 11595
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 8.467731186522007e-06,
+      "loss": 0.1151,
+      "step": 11596
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 8.467597899394878e-06,
+      "loss": 0.1056,
+      "step": 11597
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 8.467464612267749e-06,
+      "loss": 0.1623,
+      "step": 11598
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 8.467331325140618e-06,
+      "loss": 0.1182,
+      "step": 11599
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 8.467198038013489e-06,
+      "loss": 0.0945,
+      "step": 11600
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 8.46706475088636e-06,
+      "loss": 0.0824,
+      "step": 11601
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 8.466931463759231e-06,
+      "loss": 0.0881,
+      "step": 11602
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 8.466798176632102e-06,
+      "loss": 0.0907,
+      "step": 11603
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 8.466664889504971e-06,
+      "loss": 0.1238,
+      "step": 11604
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 8.466531602377844e-06,
+      "loss": 0.1569,
+      "step": 11605
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 8.466398315250713e-06,
+      "loss": 0.1073,
+      "step": 11606
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 8.466265028123584e-06,
+      "loss": 0.0954,
+      "step": 11607
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 8.466131740996455e-06,
+      "loss": 0.0839,
+      "step": 11608
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 8.465998453869326e-06,
+      "loss": 0.1478,
+      "step": 11609
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 8.465865166742197e-06,
+      "loss": 0.1096,
+      "step": 11610
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 8.465731879615068e-06,
+      "loss": 0.0927,
+      "step": 11611
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 8.465598592487937e-06,
+      "loss": 0.1201,
+      "step": 11612
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 8.46546530536081e-06,
+      "loss": 0.077,
+      "step": 11613
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 8.46533201823368e-06,
+      "loss": 0.116,
+      "step": 11614
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 8.46519873110655e-06,
+      "loss": 0.0894,
+      "step": 11615
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 8.465065443979421e-06,
+      "loss": 0.1058,
+      "step": 11616
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 8.464932156852292e-06,
+      "loss": 0.1334,
+      "step": 11617
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 8.464798869725163e-06,
+      "loss": 0.0822,
+      "step": 11618
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 8.464665582598033e-06,
+      "loss": 0.1341,
+      "step": 11619
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 8.464532295470904e-06,
+      "loss": 0.1004,
+      "step": 11620
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 8.464399008343775e-06,
+      "loss": 0.1786,
+      "step": 11621
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 8.464265721216646e-06,
+      "loss": 0.1601,
+      "step": 11622
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 8.464132434089517e-06,
+      "loss": 0.1023,
+      "step": 11623
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 8.463999146962388e-06,
+      "loss": 0.1023,
+      "step": 11624
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 8.463865859835257e-06,
+      "loss": 0.0725,
+      "step": 11625
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 8.463732572708128e-06,
+      "loss": 0.0784,
+      "step": 11626
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 8.463599285580999e-06,
+      "loss": 0.1388,
+      "step": 11627
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 8.46346599845387e-06,
+      "loss": 0.1485,
+      "step": 11628
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 8.463332711326741e-06,
+      "loss": 0.1281,
+      "step": 11629
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 8.463199424199612e-06,
+      "loss": 0.1038,
+      "step": 11630
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 8.463066137072481e-06,
+      "loss": 0.1149,
+      "step": 11631
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 8.462932849945354e-06,
+      "loss": 0.0944,
+      "step": 11632
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 8.462799562818223e-06,
+      "loss": 0.1441,
+      "step": 11633
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 8.462666275691094e-06,
+      "loss": 0.0999,
+      "step": 11634
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 8.462532988563965e-06,
+      "loss": 0.1115,
+      "step": 11635
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 8.462399701436836e-06,
+      "loss": 0.0756,
+      "step": 11636
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 8.462266414309707e-06,
+      "loss": 0.0882,
+      "step": 11637
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 8.462133127182577e-06,
+      "loss": 0.1146,
+      "step": 11638
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 8.461999840055448e-06,
+      "loss": 0.1039,
+      "step": 11639
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 8.461866552928319e-06,
+      "loss": 0.1007,
+      "step": 11640
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 8.46173326580119e-06,
+      "loss": 0.0825,
+      "step": 11641
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 8.46159997867406e-06,
+      "loss": 0.1108,
+      "step": 11642
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 8.461466691546931e-06,
+      "loss": 0.097,
+      "step": 11643
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 8.4613334044198e-06,
+      "loss": 0.1282,
+      "step": 11644
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 8.461200117292673e-06,
+      "loss": 0.0976,
+      "step": 11645
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 8.461066830165543e-06,
+      "loss": 0.1289,
+      "step": 11646
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 8.460933543038414e-06,
+      "loss": 0.0585,
+      "step": 11647
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 8.460800255911285e-06,
+      "loss": 0.1186,
+      "step": 11648
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 8.460666968784156e-06,
+      "loss": 0.0947,
+      "step": 11649
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 8.460533681657027e-06,
+      "loss": 0.1133,
+      "step": 11650
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 8.460400394529898e-06,
+      "loss": 0.0645,
+      "step": 11651
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 8.460267107402767e-06,
+      "loss": 0.1037,
+      "step": 11652
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 8.46013382027564e-06,
+      "loss": 0.1035,
+      "step": 11653
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 8.460000533148509e-06,
+      "loss": 0.0952,
+      "step": 11654
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 8.45986724602138e-06,
+      "loss": 0.0796,
+      "step": 11655
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 8.459733958894251e-06,
+      "loss": 0.1213,
+      "step": 11656
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 8.459600671767122e-06,
+      "loss": 0.1047,
+      "step": 11657
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 8.459467384639993e-06,
+      "loss": 0.1192,
+      "step": 11658
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 8.459334097512862e-06,
+      "loss": 0.1129,
+      "step": 11659
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 8.459200810385733e-06,
+      "loss": 0.0806,
+      "step": 11660
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 8.459067523258604e-06,
+      "loss": 0.1124,
+      "step": 11661
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 8.458934236131475e-06,
+      "loss": 0.1165,
+      "step": 11662
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 8.458800949004345e-06,
+      "loss": 0.1309,
+      "step": 11663
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 8.458667661877217e-06,
+      "loss": 0.0885,
+      "step": 11664
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 8.458534374750087e-06,
+      "loss": 0.0994,
+      "step": 11665
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 8.458401087622958e-06,
+      "loss": 0.0911,
+      "step": 11666
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 8.458267800495829e-06,
+      "loss": 0.0762,
+      "step": 11667
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 8.4581345133687e-06,
+      "loss": 0.1163,
+      "step": 11668
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 8.45800122624157e-06,
+      "loss": 0.1126,
+      "step": 11669
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 8.457867939114442e-06,
+      "loss": 0.1494,
+      "step": 11670
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 8.457734651987311e-06,
+      "loss": 0.1194,
+      "step": 11671
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 8.457601364860184e-06,
+      "loss": 0.14,
+      "step": 11672
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 8.457468077733053e-06,
+      "loss": 0.1656,
+      "step": 11673
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 8.457334790605924e-06,
+      "loss": 0.1043,
+      "step": 11674
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 8.457201503478795e-06,
+      "loss": 0.0604,
+      "step": 11675
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 8.457068216351666e-06,
+      "loss": 0.1185,
+      "step": 11676
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 8.456934929224537e-06,
+      "loss": 0.0946,
+      "step": 11677
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 8.456801642097406e-06,
+      "loss": 0.1552,
+      "step": 11678
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 8.456668354970277e-06,
+      "loss": 0.168,
+      "step": 11679
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 8.456535067843148e-06,
+      "loss": 0.1134,
+      "step": 11680
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 8.456401780716019e-06,
+      "loss": 0.0964,
+      "step": 11681
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 8.45626849358889e-06,
+      "loss": 0.1009,
+      "step": 11682
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 8.456135206461761e-06,
+      "loss": 0.0954,
+      "step": 11683
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 8.45600191933463e-06,
+      "loss": 0.0947,
+      "step": 11684
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 8.455868632207503e-06,
+      "loss": 0.1146,
+      "step": 11685
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 8.455735345080372e-06,
+      "loss": 0.0906,
+      "step": 11686
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 8.455602057953243e-06,
+      "loss": 0.0876,
+      "step": 11687
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 8.455468770826114e-06,
+      "loss": 0.1213,
+      "step": 11688
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 8.455335483698985e-06,
+      "loss": 0.1208,
+      "step": 11689
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 8.455202196571856e-06,
+      "loss": 0.135,
+      "step": 11690
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 8.455068909444727e-06,
+      "loss": 0.1179,
+      "step": 11691
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 8.454935622317597e-06,
+      "loss": 0.0901,
+      "step": 11692
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 8.454802335190468e-06,
+      "loss": 0.091,
+      "step": 11693
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 8.454669048063339e-06,
+      "loss": 0.0945,
+      "step": 11694
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 8.45453576093621e-06,
+      "loss": 0.0966,
+      "step": 11695
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 8.45440247380908e-06,
+      "loss": 0.107,
+      "step": 11696
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 8.45426918668195e-06,
+      "loss": 0.0847,
+      "step": 11697
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 8.454135899554821e-06,
+      "loss": 0.1242,
+      "step": 11698
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 8.454002612427692e-06,
+      "loss": 0.0889,
+      "step": 11699
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 8.453869325300563e-06,
+      "loss": 0.105,
+      "step": 11700
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 8.453736038173434e-06,
+      "loss": 0.1035,
+      "step": 11701
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 8.453602751046305e-06,
+      "loss": 0.1003,
+      "step": 11702
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 8.453469463919174e-06,
+      "loss": 0.0837,
+      "step": 11703
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 8.453336176792047e-06,
+      "loss": 0.1063,
+      "step": 11704
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 8.453202889664916e-06,
+      "loss": 0.0828,
+      "step": 11705
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 8.453069602537787e-06,
+      "loss": 0.0906,
+      "step": 11706
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 8.452936315410658e-06,
+      "loss": 0.1859,
+      "step": 11707
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 8.45280302828353e-06,
+      "loss": 0.078,
+      "step": 11708
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 8.4526697411564e-06,
+      "loss": 0.1628,
+      "step": 11709
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 8.452536454029271e-06,
+      "loss": 0.114,
+      "step": 11710
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 8.45240316690214e-06,
+      "loss": 0.1214,
+      "step": 11711
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 8.452269879775013e-06,
+      "loss": 0.0781,
+      "step": 11712
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 8.452136592647882e-06,
+      "loss": 0.1472,
+      "step": 11713
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 8.452003305520753e-06,
+      "loss": 0.0863,
+      "step": 11714
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 8.451870018393624e-06,
+      "loss": 0.0867,
+      "step": 11715
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 8.451736731266495e-06,
+      "loss": 0.1175,
+      "step": 11716
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 8.451603444139366e-06,
+      "loss": 0.0838,
+      "step": 11717
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 8.451470157012236e-06,
+      "loss": 0.08,
+      "step": 11718
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 8.451336869885107e-06,
+      "loss": 0.0941,
+      "step": 11719
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 8.451203582757978e-06,
+      "loss": 0.1141,
+      "step": 11720
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 8.451070295630849e-06,
+      "loss": 0.0969,
+      "step": 11721
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 8.45093700850372e-06,
+      "loss": 0.1092,
+      "step": 11722
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 8.45080372137659e-06,
+      "loss": 0.1041,
+      "step": 11723
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 8.45067043424946e-06,
+      "loss": 0.0656,
+      "step": 11724
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 8.450537147122331e-06,
+      "loss": 0.1101,
+      "step": 11725
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 8.450403859995202e-06,
+      "loss": 0.1409,
+      "step": 11726
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 8.450270572868073e-06,
+      "loss": 0.087,
+      "step": 11727
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 8.450137285740944e-06,
+      "loss": 0.104,
+      "step": 11728
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 8.450003998613815e-06,
+      "loss": 0.066,
+      "step": 11729
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 8.449870711486684e-06,
+      "loss": 0.0911,
+      "step": 11730
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 8.449737424359557e-06,
+      "loss": 0.1097,
+      "step": 11731
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 8.449604137232426e-06,
+      "loss": 0.0913,
+      "step": 11732
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 8.449470850105297e-06,
+      "loss": 0.0827,
+      "step": 11733
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 8.449337562978168e-06,
+      "loss": 0.1223,
+      "step": 11734
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 8.44920427585104e-06,
+      "loss": 0.1222,
+      "step": 11735
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 8.44907098872391e-06,
+      "loss": 0.087,
+      "step": 11736
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 8.44893770159678e-06,
+      "loss": 0.0954,
+      "step": 11737
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 8.44880441446965e-06,
+      "loss": 0.1082,
+      "step": 11738
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 8.448671127342522e-06,
+      "loss": 0.0904,
+      "step": 11739
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 8.448537840215393e-06,
+      "loss": 0.068,
+      "step": 11740
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 8.448404553088264e-06,
+      "loss": 0.0827,
+      "step": 11741
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 8.448271265961135e-06,
+      "loss": 0.1164,
+      "step": 11742
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 8.448137978834004e-06,
+      "loss": 0.0672,
+      "step": 11743
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 8.448004691706877e-06,
+      "loss": 0.1545,
+      "step": 11744
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 8.447871404579746e-06,
+      "loss": 0.1026,
+      "step": 11745
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 8.447738117452617e-06,
+      "loss": 0.0877,
+      "step": 11746
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 8.447604830325488e-06,
+      "loss": 0.0792,
+      "step": 11747
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 8.447471543198359e-06,
+      "loss": 0.0694,
+      "step": 11748
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 8.44733825607123e-06,
+      "loss": 0.0767,
+      "step": 11749
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 8.4472049689441e-06,
+      "loss": 0.1103,
+      "step": 11750
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 8.44707168181697e-06,
+      "loss": 0.1091,
+      "step": 11751
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 8.446938394689843e-06,
+      "loss": 0.1139,
+      "step": 11752
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 8.446805107562712e-06,
+      "loss": 0.0792,
+      "step": 11753
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 8.446671820435583e-06,
+      "loss": 0.0896,
+      "step": 11754
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 8.446538533308454e-06,
+      "loss": 0.0845,
+      "step": 11755
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 8.446405246181323e-06,
+      "loss": 0.102,
+      "step": 11756
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 8.446271959054196e-06,
+      "loss": 0.0924,
+      "step": 11757
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 8.446138671927065e-06,
+      "loss": 0.1403,
+      "step": 11758
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 8.446005384799936e-06,
+      "loss": 0.0845,
+      "step": 11759
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 8.445872097672807e-06,
+      "loss": 0.0714,
+      "step": 11760
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 8.445738810545678e-06,
+      "loss": 0.1151,
+      "step": 11761
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 8.44560552341855e-06,
+      "loss": 0.1126,
+      "step": 11762
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 8.44547223629142e-06,
+      "loss": 0.1421,
+      "step": 11763
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 8.44533894916429e-06,
+      "loss": 0.0926,
+      "step": 11764
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 8.44520566203716e-06,
+      "loss": 0.1207,
+      "step": 11765
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 8.445072374910032e-06,
+      "loss": 0.0907,
+      "step": 11766
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 8.444939087782903e-06,
+      "loss": 0.091,
+      "step": 11767
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 8.444805800655774e-06,
+      "loss": 0.085,
+      "step": 11768
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 8.444672513528645e-06,
+      "loss": 0.1149,
+      "step": 11769
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 8.444539226401514e-06,
+      "loss": 0.0913,
+      "step": 11770
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 8.444405939274387e-06,
+      "loss": 0.061,
+      "step": 11771
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 8.444272652147256e-06,
+      "loss": 0.1243,
+      "step": 11772
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 8.444139365020127e-06,
+      "loss": 0.1031,
+      "step": 11773
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 8.444006077892998e-06,
+      "loss": 0.1448,
+      "step": 11774
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 8.443872790765869e-06,
+      "loss": 0.074,
+      "step": 11775
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 8.44373950363874e-06,
+      "loss": 0.1263,
+      "step": 11776
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 8.44360621651161e-06,
+      "loss": 0.127,
+      "step": 11777
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 8.44347292938448e-06,
+      "loss": 0.0771,
+      "step": 11778
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 8.443339642257351e-06,
+      "loss": 0.1145,
+      "step": 11779
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 8.443206355130222e-06,
+      "loss": 0.1312,
+      "step": 11780
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 8.443073068003093e-06,
+      "loss": 0.1011,
+      "step": 11781
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 8.442939780875964e-06,
+      "loss": 0.096,
+      "step": 11782
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 8.442806493748833e-06,
+      "loss": 0.1222,
+      "step": 11783
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 8.442673206621706e-06,
+      "loss": 0.0937,
+      "step": 11784
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 8.442539919494575e-06,
+      "loss": 0.081,
+      "step": 11785
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 8.442406632367446e-06,
+      "loss": 0.0914,
+      "step": 11786
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 8.442273345240317e-06,
+      "loss": 0.1002,
+      "step": 11787
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 8.442140058113188e-06,
+      "loss": 0.065,
+      "step": 11788
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 8.44200677098606e-06,
+      "loss": 0.0985,
+      "step": 11789
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 8.44187348385893e-06,
+      "loss": 0.0944,
+      "step": 11790
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 8.4417401967318e-06,
+      "loss": 0.0794,
+      "step": 11791
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 8.44160690960467e-06,
+      "loss": 0.1053,
+      "step": 11792
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 8.441473622477542e-06,
+      "loss": 0.1076,
+      "step": 11793
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 8.441340335350413e-06,
+      "loss": 0.1167,
+      "step": 11794
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 8.441207048223284e-06,
+      "loss": 0.1251,
+      "step": 11795
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 8.441073761096153e-06,
+      "loss": 0.1083,
+      "step": 11796
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 8.440940473969024e-06,
+      "loss": 0.0763,
+      "step": 11797
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 8.440807186841895e-06,
+      "loss": 0.1137,
+      "step": 11798
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 8.440673899714766e-06,
+      "loss": 0.0794,
+      "step": 11799
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 8.440540612587637e-06,
+      "loss": 0.0975,
+      "step": 11800
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 8.440407325460508e-06,
+      "loss": 0.1146,
+      "step": 11801
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 8.440274038333377e-06,
+      "loss": 0.1238,
+      "step": 11802
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 8.44014075120625e-06,
+      "loss": 0.1312,
+      "step": 11803
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 8.44000746407912e-06,
+      "loss": 0.0963,
+      "step": 11804
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 8.43987417695199e-06,
+      "loss": 0.0818,
+      "step": 11805
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 8.439740889824861e-06,
+      "loss": 0.0829,
+      "step": 11806
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 8.439607602697732e-06,
+      "loss": 0.1719,
+      "step": 11807
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 8.439474315570603e-06,
+      "loss": 0.1286,
+      "step": 11808
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 8.439341028443474e-06,
+      "loss": 0.0828,
+      "step": 11809
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 8.439207741316344e-06,
+      "loss": 0.0665,
+      "step": 11810
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 8.439074454189216e-06,
+      "loss": 0.1042,
+      "step": 11811
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 8.438941167062086e-06,
+      "loss": 0.0872,
+      "step": 11812
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 8.438807879934957e-06,
+      "loss": 0.1412,
+      "step": 11813
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 8.438674592807828e-06,
+      "loss": 0.0902,
+      "step": 11814
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 8.438541305680697e-06,
+      "loss": 0.0933,
+      "step": 11815
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 8.43840801855357e-06,
+      "loss": 0.1105,
+      "step": 11816
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 8.438274731426439e-06,
+      "loss": 0.1179,
+      "step": 11817
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 8.43814144429931e-06,
+      "loss": 0.0866,
+      "step": 11818
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 8.43800815717218e-06,
+      "loss": 0.113,
+      "step": 11819
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 8.437874870045052e-06,
+      "loss": 0.0922,
+      "step": 11820
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 8.437741582917923e-06,
+      "loss": 0.0606,
+      "step": 11821
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 8.437608295790794e-06,
+      "loss": 0.1187,
+      "step": 11822
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 8.437475008663663e-06,
+      "loss": 0.0862,
+      "step": 11823
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 8.437341721536536e-06,
+      "loss": 0.1065,
+      "step": 11824
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 8.437208434409405e-06,
+      "loss": 0.0941,
+      "step": 11825
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 8.437075147282276e-06,
+      "loss": 0.1067,
+      "step": 11826
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 8.436941860155147e-06,
+      "loss": 0.1073,
+      "step": 11827
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 8.436808573028018e-06,
+      "loss": 0.0697,
+      "step": 11828
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 8.436675285900889e-06,
+      "loss": 0.1187,
+      "step": 11829
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 8.43654199877376e-06,
+      "loss": 0.1028,
+      "step": 11830
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 8.43640871164663e-06,
+      "loss": 0.0893,
+      "step": 11831
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 8.4362754245195e-06,
+      "loss": 0.1133,
+      "step": 11832
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 8.436142137392371e-06,
+      "loss": 0.0952,
+      "step": 11833
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 8.436008850265242e-06,
+      "loss": 0.0953,
+      "step": 11834
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 8.435875563138113e-06,
+      "loss": 0.0761,
+      "step": 11835
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 8.435742276010983e-06,
+      "loss": 0.1122,
+      "step": 11836
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 8.435608988883854e-06,
+      "loss": 0.1469,
+      "step": 11837
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 8.435475701756725e-06,
+      "loss": 0.1074,
+      "step": 11838
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 8.435342414629596e-06,
+      "loss": 0.0734,
+      "step": 11839
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 8.435209127502467e-06,
+      "loss": 0.0944,
+      "step": 11840
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 8.435075840375338e-06,
+      "loss": 0.0799,
+      "step": 11841
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 8.434942553248207e-06,
+      "loss": 0.09,
+      "step": 11842
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 8.43480926612108e-06,
+      "loss": 0.0986,
+      "step": 11843
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 8.434675978993949e-06,
+      "loss": 0.0868,
+      "step": 11844
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 8.43454269186682e-06,
+      "loss": 0.0638,
+      "step": 11845
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 8.434409404739691e-06,
+      "loss": 0.0992,
+      "step": 11846
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 8.434276117612562e-06,
+      "loss": 0.1095,
+      "step": 11847
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 8.434142830485433e-06,
+      "loss": 0.0599,
+      "step": 11848
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 8.434009543358304e-06,
+      "loss": 0.1184,
+      "step": 11849
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 8.433876256231173e-06,
+      "loss": 0.061,
+      "step": 11850
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 8.433742969104046e-06,
+      "loss": 0.0741,
+      "step": 11851
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 8.433609681976915e-06,
+      "loss": 0.0895,
+      "step": 11852
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 8.433476394849786e-06,
+      "loss": 0.1465,
+      "step": 11853
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 8.433343107722657e-06,
+      "loss": 0.1182,
+      "step": 11854
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 8.433209820595526e-06,
+      "loss": 0.1331,
+      "step": 11855
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 8.433076533468399e-06,
+      "loss": 0.0793,
+      "step": 11856
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 8.432943246341268e-06,
+      "loss": 0.0949,
+      "step": 11857
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 8.43280995921414e-06,
+      "loss": 0.1165,
+      "step": 11858
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 8.43267667208701e-06,
+      "loss": 0.0907,
+      "step": 11859
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 8.432543384959881e-06,
+      "loss": 0.1179,
+      "step": 11860
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 8.432410097832752e-06,
+      "loss": 0.1027,
+      "step": 11861
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 8.432276810705623e-06,
+      "loss": 0.1163,
+      "step": 11862
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 8.432143523578493e-06,
+      "loss": 0.1251,
+      "step": 11863
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 8.432010236451364e-06,
+      "loss": 0.122,
+      "step": 11864
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 8.431876949324235e-06,
+      "loss": 0.1135,
+      "step": 11865
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 8.431743662197106e-06,
+      "loss": 0.0863,
+      "step": 11866
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 8.431610375069977e-06,
+      "loss": 0.1184,
+      "step": 11867
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 8.431477087942848e-06,
+      "loss": 0.1004,
+      "step": 11868
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 8.431343800815717e-06,
+      "loss": 0.1176,
+      "step": 11869
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 8.43121051368859e-06,
+      "loss": 0.1048,
+      "step": 11870
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 8.431077226561459e-06,
+      "loss": 0.0838,
+      "step": 11871
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 8.43094393943433e-06,
+      "loss": 0.1158,
+      "step": 11872
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 8.430810652307201e-06,
+      "loss": 0.1089,
+      "step": 11873
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 8.43067736518007e-06,
+      "loss": 0.0854,
+      "step": 11874
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 8.430544078052943e-06,
+      "loss": 0.1124,
+      "step": 11875
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 8.430410790925812e-06,
+      "loss": 0.0979,
+      "step": 11876
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 8.430277503798683e-06,
+      "loss": 0.074,
+      "step": 11877
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 8.430144216671554e-06,
+      "loss": 0.0724,
+      "step": 11878
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 8.430010929544425e-06,
+      "loss": 0.0968,
+      "step": 11879
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 8.429877642417296e-06,
+      "loss": 0.102,
+      "step": 11880
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 8.429744355290167e-06,
+      "loss": 0.0821,
+      "step": 11881
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 8.429611068163037e-06,
+      "loss": 0.0771,
+      "step": 11882
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 8.42947778103591e-06,
+      "loss": 0.0917,
+      "step": 11883
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 8.429344493908779e-06,
+      "loss": 0.1132,
+      "step": 11884
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 8.42921120678165e-06,
+      "loss": 0.1021,
+      "step": 11885
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 8.42907791965452e-06,
+      "loss": 0.075,
+      "step": 11886
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 8.428944632527392e-06,
+      "loss": 0.1199,
+      "step": 11887
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 8.428811345400263e-06,
+      "loss": 0.0823,
+      "step": 11888
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 8.428678058273134e-06,
+      "loss": 0.0822,
+      "step": 11889
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 8.428544771146003e-06,
+      "loss": 0.1025,
+      "step": 11890
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 8.428411484018875e-06,
+      "loss": 0.0762,
+      "step": 11891
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 8.428278196891745e-06,
+      "loss": 0.1147,
+      "step": 11892
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 8.428144909764616e-06,
+      "loss": 0.0811,
+      "step": 11893
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 8.428011622637487e-06,
+      "loss": 0.0927,
+      "step": 11894
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 8.427878335510356e-06,
+      "loss": 0.099,
+      "step": 11895
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 8.427745048383229e-06,
+      "loss": 0.1168,
+      "step": 11896
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 8.427611761256098e-06,
+      "loss": 0.0638,
+      "step": 11897
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 8.427478474128969e-06,
+      "loss": 0.1002,
+      "step": 11898
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 8.42734518700184e-06,
+      "loss": 0.168,
+      "step": 11899
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 8.427211899874711e-06,
+      "loss": 0.0734,
+      "step": 11900
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 8.427078612747582e-06,
+      "loss": 0.11,
+      "step": 11901
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 8.426945325620453e-06,
+      "loss": 0.117,
+      "step": 11902
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 8.426812038493322e-06,
+      "loss": 0.0931,
+      "step": 11903
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 8.426678751366193e-06,
+      "loss": 0.0897,
+      "step": 11904
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 8.426545464239064e-06,
+      "loss": 0.0708,
+      "step": 11905
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 8.426412177111935e-06,
+      "loss": 0.1126,
+      "step": 11906
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 8.426278889984806e-06,
+      "loss": 0.1202,
+      "step": 11907
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 8.426145602857677e-06,
+      "loss": 0.1017,
+      "step": 11908
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 8.426012315730547e-06,
+      "loss": 0.1302,
+      "step": 11909
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 8.42587902860342e-06,
+      "loss": 0.1131,
+      "step": 11910
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 8.425745741476289e-06,
+      "loss": 0.103,
+      "step": 11911
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 8.42561245434916e-06,
+      "loss": 0.099,
+      "step": 11912
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 8.42547916722203e-06,
+      "loss": 0.1161,
+      "step": 11913
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 8.4253458800949e-06,
+      "loss": 0.1491,
+      "step": 11914
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 8.425212592967773e-06,
+      "loss": 0.1393,
+      "step": 11915
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 8.425079305840642e-06,
+      "loss": 0.0881,
+      "step": 11916
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 8.424946018713513e-06,
+      "loss": 0.1202,
+      "step": 11917
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 8.424812731586384e-06,
+      "loss": 0.0859,
+      "step": 11918
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 8.424679444459255e-06,
+      "loss": 0.1018,
+      "step": 11919
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 8.424546157332126e-06,
+      "loss": 0.0942,
+      "step": 11920
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 8.424412870204997e-06,
+      "loss": 0.1253,
+      "step": 11921
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 8.424279583077866e-06,
+      "loss": 0.1561,
+      "step": 11922
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 8.424146295950739e-06,
+      "loss": 0.1131,
+      "step": 11923
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 8.424013008823608e-06,
+      "loss": 0.1054,
+      "step": 11924
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 8.423879721696479e-06,
+      "loss": 0.0871,
+      "step": 11925
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 8.42374643456935e-06,
+      "loss": 0.0879,
+      "step": 11926
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 8.423613147442221e-06,
+      "loss": 0.0972,
+      "step": 11927
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 8.423479860315092e-06,
+      "loss": 0.1272,
+      "step": 11928
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 8.423346573187963e-06,
+      "loss": 0.0946,
+      "step": 11929
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 8.423213286060832e-06,
+      "loss": 0.1085,
+      "step": 11930
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 8.423079998933703e-06,
+      "loss": 0.1125,
+      "step": 11931
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 8.422946711806574e-06,
+      "loss": 0.0926,
+      "step": 11932
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 8.422813424679445e-06,
+      "loss": 0.0955,
+      "step": 11933
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 8.422680137552316e-06,
+      "loss": 0.1138,
+      "step": 11934
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 8.422546850425186e-06,
+      "loss": 0.084,
+      "step": 11935
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 8.422413563298057e-06,
+      "loss": 0.0845,
+      "step": 11936
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 8.422280276170928e-06,
+      "loss": 0.0732,
+      "step": 11937
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 8.422146989043799e-06,
+      "loss": 0.109,
+      "step": 11938
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 8.42201370191667e-06,
+      "loss": 0.1085,
+      "step": 11939
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 8.42188041478954e-06,
+      "loss": 0.0812,
+      "step": 11940
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 8.42174712766241e-06,
+      "loss": 0.0955,
+      "step": 11941
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 8.421613840535283e-06,
+      "loss": 0.1415,
+      "step": 11942
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 8.421480553408152e-06,
+      "loss": 0.0865,
+      "step": 11943
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 8.421347266281023e-06,
+      "loss": 0.1174,
+      "step": 11944
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 8.421213979153894e-06,
+      "loss": 0.0828,
+      "step": 11945
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 8.421080692026765e-06,
+      "loss": 0.0856,
+      "step": 11946
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 8.420947404899636e-06,
+      "loss": 0.1149,
+      "step": 11947
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 8.420814117772507e-06,
+      "loss": 0.1048,
+      "step": 11948
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 8.420680830645376e-06,
+      "loss": 0.1113,
+      "step": 11949
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 8.420547543518249e-06,
+      "loss": 0.0913,
+      "step": 11950
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 8.420414256391118e-06,
+      "loss": 0.093,
+      "step": 11951
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 8.42028096926399e-06,
+      "loss": 0.0834,
+      "step": 11952
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 8.42014768213686e-06,
+      "loss": 0.0834,
+      "step": 11953
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 8.42001439500973e-06,
+      "loss": 0.1101,
+      "step": 11954
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 8.419881107882602e-06,
+      "loss": 0.0858,
+      "step": 11955
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 8.419747820755472e-06,
+      "loss": 0.118,
+      "step": 11956
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 8.419614533628343e-06,
+      "loss": 0.129,
+      "step": 11957
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 8.419481246501214e-06,
+      "loss": 0.1123,
+      "step": 11958
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 8.419347959374085e-06,
+      "loss": 0.1088,
+      "step": 11959
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 8.419214672246955e-06,
+      "loss": 0.1025,
+      "step": 11960
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 8.419081385119826e-06,
+      "loss": 0.0755,
+      "step": 11961
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 8.418948097992696e-06,
+      "loss": 0.0934,
+      "step": 11962
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 8.418814810865567e-06,
+      "loss": 0.1313,
+      "step": 11963
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 8.418681523738438e-06,
+      "loss": 0.0756,
+      "step": 11964
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 8.418548236611309e-06,
+      "loss": 0.1012,
+      "step": 11965
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 8.41841494948418e-06,
+      "loss": 0.0921,
+      "step": 11966
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 8.41828166235705e-06,
+      "loss": 0.0875,
+      "step": 11967
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 8.41814837522992e-06,
+      "loss": 0.1205,
+      "step": 11968
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 8.418015088102793e-06,
+      "loss": 0.0788,
+      "step": 11969
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 8.417881800975662e-06,
+      "loss": 0.1128,
+      "step": 11970
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 8.417748513848533e-06,
+      "loss": 0.1073,
+      "step": 11971
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 8.417615226721404e-06,
+      "loss": 0.1425,
+      "step": 11972
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 8.417481939594273e-06,
+      "loss": 0.0688,
+      "step": 11973
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 8.417348652467146e-06,
+      "loss": 0.0961,
+      "step": 11974
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 8.417215365340015e-06,
+      "loss": 0.0796,
+      "step": 11975
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 8.417082078212886e-06,
+      "loss": 0.0957,
+      "step": 11976
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 8.416948791085757e-06,
+      "loss": 0.0883,
+      "step": 11977
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 8.416815503958628e-06,
+      "loss": 0.1082,
+      "step": 11978
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 8.4166822168315e-06,
+      "loss": 0.1163,
+      "step": 11979
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 8.41654892970437e-06,
+      "loss": 0.0856,
+      "step": 11980
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 8.41641564257724e-06,
+      "loss": 0.1126,
+      "step": 11981
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 8.416282355450112e-06,
+      "loss": 0.0837,
+      "step": 11982
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 8.416149068322982e-06,
+      "loss": 0.0701,
+      "step": 11983
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 8.416015781195853e-06,
+      "loss": 0.1047,
+      "step": 11984
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 8.415882494068724e-06,
+      "loss": 0.1072,
+      "step": 11985
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 8.415749206941595e-06,
+      "loss": 0.0958,
+      "step": 11986
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 8.415615919814466e-06,
+      "loss": 0.0873,
+      "step": 11987
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 8.415482632687337e-06,
+      "loss": 0.0978,
+      "step": 11988
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 8.415349345560206e-06,
+      "loss": 0.1024,
+      "step": 11989
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 8.415216058433079e-06,
+      "loss": 0.1057,
+      "step": 11990
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 8.415082771305948e-06,
+      "loss": 0.0981,
+      "step": 11991
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 8.414949484178819e-06,
+      "loss": 0.1093,
+      "step": 11992
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 8.41481619705169e-06,
+      "loss": 0.0898,
+      "step": 11993
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 8.414682909924559e-06,
+      "loss": 0.1176,
+      "step": 11994
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 8.414549622797432e-06,
+      "loss": 0.1016,
+      "step": 11995
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 8.414416335670301e-06,
+      "loss": 0.1053,
+      "step": 11996
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 8.414283048543172e-06,
+      "loss": 0.116,
+      "step": 11997
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 8.414149761416043e-06,
+      "loss": 0.0972,
+      "step": 11998
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 8.414016474288914e-06,
+      "loss": 0.1035,
+      "step": 11999
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 8.413883187161785e-06,
+      "loss": 0.0693,
+      "step": 12000
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 8.413749900034656e-06,
+      "loss": 0.1176,
+      "step": 12001
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 8.413616612907525e-06,
+      "loss": 0.1127,
+      "step": 12002
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 8.413483325780396e-06,
+      "loss": 0.0793,
+      "step": 12003
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 8.413350038653267e-06,
+      "loss": 0.0818,
+      "step": 12004
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 8.413216751526138e-06,
+      "loss": 0.1439,
+      "step": 12005
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 8.41308346439901e-06,
+      "loss": 0.0765,
+      "step": 12006
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 8.41295017727188e-06,
+      "loss": 0.1474,
+      "step": 12007
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 8.41281689014475e-06,
+      "loss": 0.1109,
+      "step": 12008
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 8.412683603017622e-06,
+      "loss": 0.0698,
+      "step": 12009
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 8.412550315890492e-06,
+      "loss": 0.0952,
+      "step": 12010
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 8.412417028763363e-06,
+      "loss": 0.1375,
+      "step": 12011
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 8.412283741636234e-06,
+      "loss": 0.1062,
+      "step": 12012
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 8.412150454509103e-06,
+      "loss": 0.1116,
+      "step": 12013
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 8.412017167381976e-06,
+      "loss": 0.1042,
+      "step": 12014
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 8.411883880254845e-06,
+      "loss": 0.0882,
+      "step": 12015
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 8.411750593127716e-06,
+      "loss": 0.1302,
+      "step": 12016
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 8.411617306000587e-06,
+      "loss": 0.1135,
+      "step": 12017
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 8.411484018873458e-06,
+      "loss": 0.0822,
+      "step": 12018
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 8.411350731746329e-06,
+      "loss": 0.1523,
+      "step": 12019
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 8.4112174446192e-06,
+      "loss": 0.1135,
+      "step": 12020
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 8.41108415749207e-06,
+      "loss": 0.0883,
+      "step": 12021
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 8.410950870364942e-06,
+      "loss": 0.1324,
+      "step": 12022
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 8.410817583237811e-06,
+      "loss": 0.083,
+      "step": 12023
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 8.410684296110682e-06,
+      "loss": 0.0877,
+      "step": 12024
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 8.410551008983553e-06,
+      "loss": 0.0897,
+      "step": 12025
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 8.410417721856424e-06,
+      "loss": 0.079,
+      "step": 12026
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 8.410284434729295e-06,
+      "loss": 0.0621,
+      "step": 12027
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 8.410151147602166e-06,
+      "loss": 0.1582,
+      "step": 12028
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 8.410017860475035e-06,
+      "loss": 0.1095,
+      "step": 12029
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 8.409884573347906e-06,
+      "loss": 0.0648,
+      "step": 12030
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 8.409751286220777e-06,
+      "loss": 0.1099,
+      "step": 12031
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 8.409617999093648e-06,
+      "loss": 0.1222,
+      "step": 12032
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 8.40948471196652e-06,
+      "loss": 0.1832,
+      "step": 12033
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 8.409351424839389e-06,
+      "loss": 0.1118,
+      "step": 12034
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 8.40921813771226e-06,
+      "loss": 0.1392,
+      "step": 12035
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 8.40908485058513e-06,
+      "loss": 0.1044,
+      "step": 12036
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 8.408951563458002e-06,
+      "loss": 0.1031,
+      "step": 12037
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 8.408818276330873e-06,
+      "loss": 0.1554,
+      "step": 12038
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 8.408684989203744e-06,
+      "loss": 0.0902,
+      "step": 12039
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 8.408551702076613e-06,
+      "loss": 0.0823,
+      "step": 12040
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 8.408418414949486e-06,
+      "loss": 0.0722,
+      "step": 12041
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 8.408285127822355e-06,
+      "loss": 0.0737,
+      "step": 12042
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 8.408151840695226e-06,
+      "loss": 0.1008,
+      "step": 12043
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 8.408018553568097e-06,
+      "loss": 0.109,
+      "step": 12044
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 8.407885266440968e-06,
+      "loss": 0.1031,
+      "step": 12045
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 8.407751979313839e-06,
+      "loss": 0.1502,
+      "step": 12046
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 8.40761869218671e-06,
+      "loss": 0.1587,
+      "step": 12047
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 8.40748540505958e-06,
+      "loss": 0.0724,
+      "step": 12048
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 8.407352117932452e-06,
+      "loss": 0.0788,
+      "step": 12049
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 8.407218830805321e-06,
+      "loss": 0.0954,
+      "step": 12050
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 8.407085543678192e-06,
+      "loss": 0.1118,
+      "step": 12051
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 8.406952256551063e-06,
+      "loss": 0.0797,
+      "step": 12052
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 8.406818969423933e-06,
+      "loss": 0.1315,
+      "step": 12053
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 8.406685682296805e-06,
+      "loss": 0.0819,
+      "step": 12054
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 8.406552395169675e-06,
+      "loss": 0.0904,
+      "step": 12055
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 8.406419108042546e-06,
+      "loss": 0.1366,
+      "step": 12056
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 8.406285820915417e-06,
+      "loss": 0.1126,
+      "step": 12057
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 8.406152533788288e-06,
+      "loss": 0.0903,
+      "step": 12058
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 8.406019246661159e-06,
+      "loss": 0.09,
+      "step": 12059
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 8.40588595953403e-06,
+      "loss": 0.0515,
+      "step": 12060
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 8.405752672406899e-06,
+      "loss": 0.0826,
+      "step": 12061
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 8.405619385279772e-06,
+      "loss": 0.0953,
+      "step": 12062
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 8.40548609815264e-06,
+      "loss": 0.1165,
+      "step": 12063
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 8.405352811025512e-06,
+      "loss": 0.1044,
+      "step": 12064
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 8.405219523898383e-06,
+      "loss": 0.1253,
+      "step": 12065
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 8.405086236771254e-06,
+      "loss": 0.1148,
+      "step": 12066
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 8.404952949644125e-06,
+      "loss": 0.1145,
+      "step": 12067
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 8.404819662516996e-06,
+      "loss": 0.1211,
+      "step": 12068
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 8.404686375389865e-06,
+      "loss": 0.124,
+      "step": 12069
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 8.404553088262736e-06,
+      "loss": 0.0904,
+      "step": 12070
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 8.404419801135607e-06,
+      "loss": 0.0886,
+      "step": 12071
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 8.404286514008478e-06,
+      "loss": 0.1292,
+      "step": 12072
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 8.404153226881349e-06,
+      "loss": 0.1082,
+      "step": 12073
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 8.404019939754218e-06,
+      "loss": 0.1132,
+      "step": 12074
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 8.40388665262709e-06,
+      "loss": 0.0946,
+      "step": 12075
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 8.40375336549996e-06,
+      "loss": 0.0777,
+      "step": 12076
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 8.403620078372831e-06,
+      "loss": 0.1155,
+      "step": 12077
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 8.403486791245702e-06,
+      "loss": 0.088,
+      "step": 12078
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 8.403353504118573e-06,
+      "loss": 0.0989,
+      "step": 12079
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 8.403220216991443e-06,
+      "loss": 0.0634,
+      "step": 12080
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 8.403086929864315e-06,
+      "loss": 0.0984,
+      "step": 12081
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 8.402953642737185e-06,
+      "loss": 0.0622,
+      "step": 12082
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 8.402820355610056e-06,
+      "loss": 0.1,
+      "step": 12083
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 8.402687068482927e-06,
+      "loss": 0.1419,
+      "step": 12084
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 8.402553781355798e-06,
+      "loss": 0.0912,
+      "step": 12085
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 8.402420494228669e-06,
+      "loss": 0.072,
+      "step": 12086
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 8.40228720710154e-06,
+      "loss": 0.1043,
+      "step": 12087
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 8.402153919974409e-06,
+      "loss": 0.1039,
+      "step": 12088
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 8.40202063284728e-06,
+      "loss": 0.0904,
+      "step": 12089
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 8.401887345720151e-06,
+      "loss": 0.0935,
+      "step": 12090
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 8.401754058593022e-06,
+      "loss": 0.0812,
+      "step": 12091
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 8.401620771465893e-06,
+      "loss": 0.1069,
+      "step": 12092
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 8.401487484338762e-06,
+      "loss": 0.0934,
+      "step": 12093
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 8.401354197211635e-06,
+      "loss": 0.1042,
+      "step": 12094
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 8.401220910084504e-06,
+      "loss": 0.1108,
+      "step": 12095
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 8.401087622957375e-06,
+      "loss": 0.1016,
+      "step": 12096
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 8.400954335830246e-06,
+      "loss": 0.093,
+      "step": 12097
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 8.400821048703117e-06,
+      "loss": 0.1193,
+      "step": 12098
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 8.400687761575988e-06,
+      "loss": 0.123,
+      "step": 12099
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 8.40055447444886e-06,
+      "loss": 0.1128,
+      "step": 12100
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 8.400421187321728e-06,
+      "loss": 0.086,
+      "step": 12101
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 8.4002879001946e-06,
+      "loss": 0.1336,
+      "step": 12102
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 8.40015461306747e-06,
+      "loss": 0.1128,
+      "step": 12103
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 8.400021325940341e-06,
+      "loss": 0.1399,
+      "step": 12104
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 8.399888038813212e-06,
+      "loss": 0.0823,
+      "step": 12105
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 8.399754751686083e-06,
+      "loss": 0.0754,
+      "step": 12106
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 8.399621464558953e-06,
+      "loss": 0.0783,
+      "step": 12107
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 8.399488177431825e-06,
+      "loss": 0.0586,
+      "step": 12108
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 8.399354890304695e-06,
+      "loss": 0.075,
+      "step": 12109
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 8.399221603177566e-06,
+      "loss": 0.104,
+      "step": 12110
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 8.399088316050437e-06,
+      "loss": 0.1008,
+      "step": 12111
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 8.398955028923306e-06,
+      "loss": 0.089,
+      "step": 12112
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 8.398821741796179e-06,
+      "loss": 0.1407,
+      "step": 12113
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 8.398688454669048e-06,
+      "loss": 0.1072,
+      "step": 12114
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 8.398555167541919e-06,
+      "loss": 0.1421,
+      "step": 12115
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 8.39842188041479e-06,
+      "loss": 0.1589,
+      "step": 12116
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 8.398288593287661e-06,
+      "loss": 0.0952,
+      "step": 12117
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 8.398155306160532e-06,
+      "loss": 0.1379,
+      "step": 12118
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 8.398022019033403e-06,
+      "loss": 0.1012,
+      "step": 12119
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 8.397888731906272e-06,
+      "loss": 0.1079,
+      "step": 12120
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 8.397755444779145e-06,
+      "loss": 0.1152,
+      "step": 12121
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 8.397622157652014e-06,
+      "loss": 0.1303,
+      "step": 12122
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 8.397488870524885e-06,
+      "loss": 0.1032,
+      "step": 12123
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 8.397355583397756e-06,
+      "loss": 0.1111,
+      "step": 12124
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 8.397222296270627e-06,
+      "loss": 0.1008,
+      "step": 12125
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 8.397089009143498e-06,
+      "loss": 0.1234,
+      "step": 12126
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 8.39695572201637e-06,
+      "loss": 0.1145,
+      "step": 12127
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 8.396822434889239e-06,
+      "loss": 0.0512,
+      "step": 12128
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 8.39668914776211e-06,
+      "loss": 0.1162,
+      "step": 12129
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 8.39655586063498e-06,
+      "loss": 0.0874,
+      "step": 12130
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 8.396422573507852e-06,
+      "loss": 0.1054,
+      "step": 12131
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 8.396289286380723e-06,
+      "loss": 0.1383,
+      "step": 12132
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 8.396155999253592e-06,
+      "loss": 0.097,
+      "step": 12133
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 8.396022712126465e-06,
+      "loss": 0.0938,
+      "step": 12134
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 8.395889424999334e-06,
+      "loss": 0.0962,
+      "step": 12135
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 8.395756137872205e-06,
+      "loss": 0.1582,
+      "step": 12136
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 8.395622850745076e-06,
+      "loss": 0.0542,
+      "step": 12137
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 8.395489563617947e-06,
+      "loss": 0.1161,
+      "step": 12138
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 8.395356276490818e-06,
+      "loss": 0.1351,
+      "step": 12139
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 8.395222989363689e-06,
+      "loss": 0.1118,
+      "step": 12140
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 8.395089702236558e-06,
+      "loss": 0.1036,
+      "step": 12141
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 8.394956415109429e-06,
+      "loss": 0.1356,
+      "step": 12142
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 8.3948231279823e-06,
+      "loss": 0.1037,
+      "step": 12143
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 8.394689840855171e-06,
+      "loss": 0.0774,
+      "step": 12144
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 8.394556553728042e-06,
+      "loss": 0.1274,
+      "step": 12145
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 8.394423266600913e-06,
+      "loss": 0.0959,
+      "step": 12146
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 8.394289979473782e-06,
+      "loss": 0.1296,
+      "step": 12147
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 8.394156692346653e-06,
+      "loss": 0.0998,
+      "step": 12148
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 8.394023405219524e-06,
+      "loss": 0.0827,
+      "step": 12149
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 8.393890118092395e-06,
+      "loss": 0.1082,
+      "step": 12150
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 8.393756830965266e-06,
+      "loss": 0.1164,
+      "step": 12151
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 8.393623543838136e-06,
+      "loss": 0.088,
+      "step": 12152
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 8.393490256711008e-06,
+      "loss": 0.1139,
+      "step": 12153
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 8.393356969583878e-06,
+      "loss": 0.0899,
+      "step": 12154
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 8.393223682456749e-06,
+      "loss": 0.0665,
+      "step": 12155
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 8.39309039532962e-06,
+      "loss": 0.0803,
+      "step": 12156
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 8.39295710820249e-06,
+      "loss": 0.1081,
+      "step": 12157
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 8.392823821075362e-06,
+      "loss": 0.1371,
+      "step": 12158
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 8.392690533948233e-06,
+      "loss": 0.1019,
+      "step": 12159
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 8.392557246821102e-06,
+      "loss": 0.0694,
+      "step": 12160
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 8.392423959693975e-06,
+      "loss": 0.1194,
+      "step": 12161
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 8.392290672566844e-06,
+      "loss": 0.1065,
+      "step": 12162
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 8.392157385439715e-06,
+      "loss": 0.1462,
+      "step": 12163
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 8.392024098312586e-06,
+      "loss": 0.1279,
+      "step": 12164
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 8.391890811185457e-06,
+      "loss": 0.1068,
+      "step": 12165
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 8.391757524058328e-06,
+      "loss": 0.0756,
+      "step": 12166
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 8.391624236931197e-06,
+      "loss": 0.0724,
+      "step": 12167
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 8.391490949804068e-06,
+      "loss": 0.102,
+      "step": 12168
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 8.39135766267694e-06,
+      "loss": 0.1339,
+      "step": 12169
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 8.39122437554981e-06,
+      "loss": 0.093,
+      "step": 12170
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 8.391091088422681e-06,
+      "loss": 0.1,
+      "step": 12171
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 8.390957801295552e-06,
+      "loss": 0.1223,
+      "step": 12172
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 8.390824514168421e-06,
+      "loss": 0.1096,
+      "step": 12173
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 8.390691227041292e-06,
+      "loss": 0.1053,
+      "step": 12174
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 8.390557939914163e-06,
+      "loss": 0.0808,
+      "step": 12175
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 8.390424652787034e-06,
+      "loss": 0.0897,
+      "step": 12176
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 8.390291365659905e-06,
+      "loss": 0.0937,
+      "step": 12177
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 8.390158078532776e-06,
+      "loss": 0.0816,
+      "step": 12178
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 8.390024791405646e-06,
+      "loss": 0.1009,
+      "step": 12179
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 8.389891504278518e-06,
+      "loss": 0.1044,
+      "step": 12180
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 8.389758217151388e-06,
+      "loss": 0.0975,
+      "step": 12181
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 8.389624930024259e-06,
+      "loss": 0.066,
+      "step": 12182
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 8.38949164289713e-06,
+      "loss": 0.0959,
+      "step": 12183
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 8.38935835577e-06,
+      "loss": 0.0941,
+      "step": 12184
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 8.389225068642872e-06,
+      "loss": 0.1292,
+      "step": 12185
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 8.389091781515743e-06,
+      "loss": 0.1328,
+      "step": 12186
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 8.388958494388612e-06,
+      "loss": 0.0774,
+      "step": 12187
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 8.388825207261483e-06,
+      "loss": 0.0818,
+      "step": 12188
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 8.388691920134354e-06,
+      "loss": 0.104,
+      "step": 12189
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 8.388558633007225e-06,
+      "loss": 0.1087,
+      "step": 12190
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 8.388425345880096e-06,
+      "loss": 0.1297,
+      "step": 12191
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 8.388292058752965e-06,
+      "loss": 0.1036,
+      "step": 12192
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 8.388158771625838e-06,
+      "loss": 0.0861,
+      "step": 12193
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 8.388025484498707e-06,
+      "loss": 0.1102,
+      "step": 12194
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 8.387892197371578e-06,
+      "loss": 0.1022,
+      "step": 12195
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 8.38775891024445e-06,
+      "loss": 0.1319,
+      "step": 12196
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 8.38762562311732e-06,
+      "loss": 0.1208,
+      "step": 12197
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 8.387492335990191e-06,
+      "loss": 0.1031,
+      "step": 12198
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 8.387359048863062e-06,
+      "loss": 0.0932,
+      "step": 12199
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 8.387225761735932e-06,
+      "loss": 0.0842,
+      "step": 12200
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 8.387092474608803e-06,
+      "loss": 0.1125,
+      "step": 12201
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 8.386959187481674e-06,
+      "loss": 0.1318,
+      "step": 12202
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 8.386825900354545e-06,
+      "loss": 0.1073,
+      "step": 12203
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 8.386692613227416e-06,
+      "loss": 0.1191,
+      "step": 12204
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 8.386559326100287e-06,
+      "loss": 0.1334,
+      "step": 12205
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 8.386426038973156e-06,
+      "loss": 0.0992,
+      "step": 12206
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 8.386292751846027e-06,
+      "loss": 0.1271,
+      "step": 12207
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 8.386159464718898e-06,
+      "loss": 0.1083,
+      "step": 12208
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 8.386026177591769e-06,
+      "loss": 0.1074,
+      "step": 12209
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 8.38589289046464e-06,
+      "loss": 0.1129,
+      "step": 12210
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 8.385759603337509e-06,
+      "loss": 0.1248,
+      "step": 12211
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 8.385626316210382e-06,
+      "loss": 0.0998,
+      "step": 12212
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 8.385493029083251e-06,
+      "loss": 0.1011,
+      "step": 12213
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 8.385359741956122e-06,
+      "loss": 0.1174,
+      "step": 12214
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 8.385226454828993e-06,
+      "loss": 0.1044,
+      "step": 12215
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 8.385093167701864e-06,
+      "loss": 0.1122,
+      "step": 12216
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 8.384959880574735e-06,
+      "loss": 0.1373,
+      "step": 12217
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 8.384826593447606e-06,
+      "loss": 0.1194,
+      "step": 12218
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 8.384693306320475e-06,
+      "loss": 0.112,
+      "step": 12219
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 8.384560019193348e-06,
+      "loss": 0.1042,
+      "step": 12220
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 8.384426732066217e-06,
+      "loss": 0.1437,
+      "step": 12221
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 8.384293444939088e-06,
+      "loss": 0.0981,
+      "step": 12222
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 8.38416015781196e-06,
+      "loss": 0.1246,
+      "step": 12223
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 8.38402687068483e-06,
+      "loss": 0.074,
+      "step": 12224
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 8.383893583557701e-06,
+      "loss": 0.1375,
+      "step": 12225
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 8.38376029643057e-06,
+      "loss": 0.126,
+      "step": 12226
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 8.383627009303442e-06,
+      "loss": 0.0807,
+      "step": 12227
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 8.383493722176313e-06,
+      "loss": 0.0777,
+      "step": 12228
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 8.383360435049184e-06,
+      "loss": 0.1587,
+      "step": 12229
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 8.383227147922055e-06,
+      "loss": 0.1243,
+      "step": 12230
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 8.383093860794926e-06,
+      "loss": 0.0996,
+      "step": 12231
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 8.382960573667795e-06,
+      "loss": 0.1199,
+      "step": 12232
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 8.382827286540668e-06,
+      "loss": 0.0718,
+      "step": 12233
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 8.382693999413537e-06,
+      "loss": 0.1128,
+      "step": 12234
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 8.382560712286408e-06,
+      "loss": 0.1418,
+      "step": 12235
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 8.382427425159279e-06,
+      "loss": 0.1182,
+      "step": 12236
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 8.38229413803215e-06,
+      "loss": 0.0988,
+      "step": 12237
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 8.382160850905021e-06,
+      "loss": 0.1017,
+      "step": 12238
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 8.382027563777892e-06,
+      "loss": 0.1274,
+      "step": 12239
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 8.381894276650761e-06,
+      "loss": 0.1178,
+      "step": 12240
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 8.381760989523632e-06,
+      "loss": 0.0851,
+      "step": 12241
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 8.381627702396503e-06,
+      "loss": 0.0903,
+      "step": 12242
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 8.381494415269374e-06,
+      "loss": 0.0774,
+      "step": 12243
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 8.381361128142245e-06,
+      "loss": 0.1164,
+      "step": 12244
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 8.381227841015116e-06,
+      "loss": 0.1091,
+      "step": 12245
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 8.381094553887985e-06,
+      "loss": 0.0883,
+      "step": 12246
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 8.380961266760856e-06,
+      "loss": 0.0873,
+      "step": 12247
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 8.380827979633727e-06,
+      "loss": 0.1079,
+      "step": 12248
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 8.380694692506598e-06,
+      "loss": 0.0828,
+      "step": 12249
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 8.38056140537947e-06,
+      "loss": 0.0987,
+      "step": 12250
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 8.380428118252339e-06,
+      "loss": 0.0937,
+      "step": 12251
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 8.380294831125211e-06,
+      "loss": 0.0924,
+      "step": 12252
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 8.38016154399808e-06,
+      "loss": 0.0706,
+      "step": 12253
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 8.380028256870952e-06,
+      "loss": 0.1391,
+      "step": 12254
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 8.379894969743823e-06,
+      "loss": 0.0813,
+      "step": 12255
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 8.379761682616694e-06,
+      "loss": 0.0779,
+      "step": 12256
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 8.379628395489565e-06,
+      "loss": 0.0916,
+      "step": 12257
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 8.379495108362436e-06,
+      "loss": 0.0932,
+      "step": 12258
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 8.379361821235305e-06,
+      "loss": 0.0793,
+      "step": 12259
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 8.379228534108178e-06,
+      "loss": 0.1278,
+      "step": 12260
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 8.379095246981047e-06,
+      "loss": 0.1106,
+      "step": 12261
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 8.378961959853918e-06,
+      "loss": 0.0743,
+      "step": 12262
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 8.378828672726789e-06,
+      "loss": 0.0803,
+      "step": 12263
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 8.37869538559966e-06,
+      "loss": 0.0838,
+      "step": 12264
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 8.378562098472531e-06,
+      "loss": 0.1003,
+      "step": 12265
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 8.3784288113454e-06,
+      "loss": 0.1156,
+      "step": 12266
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 8.378295524218271e-06,
+      "loss": 0.1259,
+      "step": 12267
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 8.378162237091142e-06,
+      "loss": 0.1059,
+      "step": 12268
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 8.378028949964013e-06,
+      "loss": 0.111,
+      "step": 12269
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 8.377895662836884e-06,
+      "loss": 0.0823,
+      "step": 12270
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 8.377762375709755e-06,
+      "loss": 0.0879,
+      "step": 12271
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 8.377629088582625e-06,
+      "loss": 0.0931,
+      "step": 12272
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 8.377495801455496e-06,
+      "loss": 0.0948,
+      "step": 12273
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 8.377362514328367e-06,
+      "loss": 0.0942,
+      "step": 12274
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 8.377229227201238e-06,
+      "loss": 0.0944,
+      "step": 12275
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 8.377095940074109e-06,
+      "loss": 0.1057,
+      "step": 12276
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 8.37696265294698e-06,
+      "loss": 0.1154,
+      "step": 12277
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 8.376829365819849e-06,
+      "loss": 0.1614,
+      "step": 12278
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 8.376696078692721e-06,
+      "loss": 0.1241,
+      "step": 12279
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 8.37656279156559e-06,
+      "loss": 0.0889,
+      "step": 12280
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 8.376429504438462e-06,
+      "loss": 0.1152,
+      "step": 12281
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 8.376296217311333e-06,
+      "loss": 0.0909,
+      "step": 12282
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 8.376162930184204e-06,
+      "loss": 0.0796,
+      "step": 12283
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 8.376029643057075e-06,
+      "loss": 0.107,
+      "step": 12284
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 8.375896355929944e-06,
+      "loss": 0.0977,
+      "step": 12285
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 8.375763068802815e-06,
+      "loss": 0.1011,
+      "step": 12286
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 8.375629781675686e-06,
+      "loss": 0.0833,
+      "step": 12287
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 8.375496494548557e-06,
+      "loss": 0.1179,
+      "step": 12288
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 8.375363207421428e-06,
+      "loss": 0.0691,
+      "step": 12289
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 8.375229920294299e-06,
+      "loss": 0.0746,
+      "step": 12290
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 8.375096633167168e-06,
+      "loss": 0.0865,
+      "step": 12291
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 8.374963346040041e-06,
+      "loss": 0.0877,
+      "step": 12292
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 8.37483005891291e-06,
+      "loss": 0.0871,
+      "step": 12293
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 8.374696771785781e-06,
+      "loss": 0.0999,
+      "step": 12294
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 8.374563484658652e-06,
+      "loss": 0.1397,
+      "step": 12295
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 8.374430197531523e-06,
+      "loss": 0.0986,
+      "step": 12296
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 8.374296910404394e-06,
+      "loss": 0.102,
+      "step": 12297
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 8.374163623277265e-06,
+      "loss": 0.0963,
+      "step": 12298
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 8.374030336150135e-06,
+      "loss": 0.0859,
+      "step": 12299
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 8.373897049023007e-06,
+      "loss": 0.1383,
+      "step": 12300
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 8.373763761895877e-06,
+      "loss": 0.1196,
+      "step": 12301
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 8.373630474768748e-06,
+      "loss": 0.1086,
+      "step": 12302
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 8.373497187641619e-06,
+      "loss": 0.0933,
+      "step": 12303
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 8.373363900514488e-06,
+      "loss": 0.1247,
+      "step": 12304
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 8.37323061338736e-06,
+      "loss": 0.1095,
+      "step": 12305
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 8.37309732626023e-06,
+      "loss": 0.0885,
+      "step": 12306
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 8.372964039133101e-06,
+      "loss": 0.1078,
+      "step": 12307
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 8.372830752005972e-06,
+      "loss": 0.1078,
+      "step": 12308
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 8.372697464878843e-06,
+      "loss": 0.0848,
+      "step": 12309
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 8.372564177751714e-06,
+      "loss": 0.0725,
+      "step": 12310
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 8.372430890624585e-06,
+      "loss": 0.0887,
+      "step": 12311
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 8.372297603497454e-06,
+      "loss": 0.0935,
+      "step": 12312
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 8.372164316370325e-06,
+      "loss": 0.0756,
+      "step": 12313
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 8.372031029243196e-06,
+      "loss": 0.1102,
+      "step": 12314
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 8.371897742116067e-06,
+      "loss": 0.0892,
+      "step": 12315
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 8.371764454988938e-06,
+      "loss": 0.0917,
+      "step": 12316
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 8.371631167861809e-06,
+      "loss": 0.0607,
+      "step": 12317
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 8.371497880734678e-06,
+      "loss": 0.1187,
+      "step": 12318
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 8.371364593607551e-06,
+      "loss": 0.1078,
+      "step": 12319
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 8.37123130648042e-06,
+      "loss": 0.0891,
+      "step": 12320
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 8.371098019353291e-06,
+      "loss": 0.0885,
+      "step": 12321
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 8.370964732226162e-06,
+      "loss": 0.1259,
+      "step": 12322
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 8.370831445099033e-06,
+      "loss": 0.1174,
+      "step": 12323
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 8.370698157971904e-06,
+      "loss": 0.1248,
+      "step": 12324
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 8.370564870844774e-06,
+      "loss": 0.0979,
+      "step": 12325
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 8.370431583717645e-06,
+      "loss": 0.1064,
+      "step": 12326
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 8.370298296590516e-06,
+      "loss": 0.0982,
+      "step": 12327
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 8.370165009463387e-06,
+      "loss": 0.1229,
+      "step": 12328
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 8.370031722336258e-06,
+      "loss": 0.0706,
+      "step": 12329
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 8.369898435209129e-06,
+      "loss": 0.1073,
+      "step": 12330
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 8.369765148081998e-06,
+      "loss": 0.0878,
+      "step": 12331
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 8.36963186095487e-06,
+      "loss": 0.1069,
+      "step": 12332
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 8.36949857382774e-06,
+      "loss": 0.1033,
+      "step": 12333
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 8.369365286700611e-06,
+      "loss": 0.0794,
+      "step": 12334
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 8.369231999573482e-06,
+      "loss": 0.1042,
+      "step": 12335
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 8.369098712446353e-06,
+      "loss": 0.0989,
+      "step": 12336
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 8.368965425319224e-06,
+      "loss": 0.1227,
+      "step": 12337
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 8.368832138192095e-06,
+      "loss": 0.0849,
+      "step": 12338
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 8.368698851064964e-06,
+      "loss": 0.0913,
+      "step": 12339
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 8.368565563937835e-06,
+      "loss": 0.0863,
+      "step": 12340
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 8.368432276810706e-06,
+      "loss": 0.0631,
+      "step": 12341
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 8.368298989683577e-06,
+      "loss": 0.0954,
+      "step": 12342
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 8.368165702556448e-06,
+      "loss": 0.0852,
+      "step": 12343
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 8.368032415429318e-06,
+      "loss": 0.0891,
+      "step": 12344
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 8.367899128302189e-06,
+      "loss": 0.0766,
+      "step": 12345
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 8.36776584117506e-06,
+      "loss": 0.1193,
+      "step": 12346
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 8.36763255404793e-06,
+      "loss": 0.0913,
+      "step": 12347
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 8.367499266920801e-06,
+      "loss": 0.1228,
+      "step": 12348
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 8.367365979793672e-06,
+      "loss": 0.1102,
+      "step": 12349
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 8.367232692666542e-06,
+      "loss": 0.072,
+      "step": 12350
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 8.367099405539414e-06,
+      "loss": 0.1092,
+      "step": 12351
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 8.366966118412284e-06,
+      "loss": 0.1284,
+      "step": 12352
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 8.366832831285155e-06,
+      "loss": 0.1008,
+      "step": 12353
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 8.366699544158026e-06,
+      "loss": 0.0986,
+      "step": 12354
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 8.366566257030897e-06,
+      "loss": 0.1061,
+      "step": 12355
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 8.366432969903768e-06,
+      "loss": 0.1266,
+      "step": 12356
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 8.366299682776639e-06,
+      "loss": 0.0934,
+      "step": 12357
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 8.366166395649508e-06,
+      "loss": 0.106,
+      "step": 12358
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 8.36603310852238e-06,
+      "loss": 0.1121,
+      "step": 12359
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 8.36589982139525e-06,
+      "loss": 0.0798,
+      "step": 12360
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 8.365766534268121e-06,
+      "loss": 0.0804,
+      "step": 12361
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 8.365633247140992e-06,
+      "loss": 0.1189,
+      "step": 12362
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 8.365499960013861e-06,
+      "loss": 0.0764,
+      "step": 12363
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 8.365366672886734e-06,
+      "loss": 0.1167,
+      "step": 12364
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 8.365233385759603e-06,
+      "loss": 0.1106,
+      "step": 12365
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 8.365100098632474e-06,
+      "loss": 0.1258,
+      "step": 12366
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 8.364966811505345e-06,
+      "loss": 0.1446,
+      "step": 12367
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 8.364833524378216e-06,
+      "loss": 0.075,
+      "step": 12368
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 8.364700237251087e-06,
+      "loss": 0.1019,
+      "step": 12369
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 8.364566950123958e-06,
+      "loss": 0.125,
+      "step": 12370
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 8.364433662996828e-06,
+      "loss": 0.0739,
+      "step": 12371
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 8.3643003758697e-06,
+      "loss": 0.0645,
+      "step": 12372
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 8.36416708874257e-06,
+      "loss": 0.0947,
+      "step": 12373
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 8.36403380161544e-06,
+      "loss": 0.0919,
+      "step": 12374
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 8.363900514488312e-06,
+      "loss": 0.0885,
+      "step": 12375
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 8.363767227361183e-06,
+      "loss": 0.0688,
+      "step": 12376
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 8.363633940234054e-06,
+      "loss": 0.1027,
+      "step": 12377
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 8.363500653106925e-06,
+      "loss": 0.0849,
+      "step": 12378
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 8.363367365979794e-06,
+      "loss": 0.0878,
+      "step": 12379
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 8.363234078852665e-06,
+      "loss": 0.0755,
+      "step": 12380
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 8.363100791725536e-06,
+      "loss": 0.0947,
+      "step": 12381
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 8.362967504598407e-06,
+      "loss": 0.1054,
+      "step": 12382
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 8.362834217471278e-06,
+      "loss": 0.092,
+      "step": 12383
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 8.362700930344147e-06,
+      "loss": 0.1724,
+      "step": 12384
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 8.362567643217018e-06,
+      "loss": 0.1096,
+      "step": 12385
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 8.362434356089889e-06,
+      "loss": 0.1348,
+      "step": 12386
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 8.36230106896276e-06,
+      "loss": 0.0565,
+      "step": 12387
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 8.362167781835631e-06,
+      "loss": 0.113,
+      "step": 12388
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 8.362034494708502e-06,
+      "loss": 0.1012,
+      "step": 12389
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 8.361901207581371e-06,
+      "loss": 0.1248,
+      "step": 12390
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 8.361767920454244e-06,
+      "loss": 0.1288,
+      "step": 12391
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 8.361634633327113e-06,
+      "loss": 0.1378,
+      "step": 12392
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 8.361501346199984e-06,
+      "loss": 0.107,
+      "step": 12393
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 8.361368059072855e-06,
+      "loss": 0.1106,
+      "step": 12394
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 8.361234771945726e-06,
+      "loss": 0.1468,
+      "step": 12395
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 8.361101484818597e-06,
+      "loss": 0.1081,
+      "step": 12396
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 8.360968197691468e-06,
+      "loss": 0.1153,
+      "step": 12397
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 8.360834910564338e-06,
+      "loss": 0.0716,
+      "step": 12398
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 8.36070162343721e-06,
+      "loss": 0.0907,
+      "step": 12399
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 8.36056833631008e-06,
+      "loss": 0.0847,
+      "step": 12400
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 8.36043504918295e-06,
+      "loss": 0.1265,
+      "step": 12401
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 8.360301762055822e-06,
+      "loss": 0.0983,
+      "step": 12402
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 8.360168474928691e-06,
+      "loss": 0.1381,
+      "step": 12403
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 8.360035187801564e-06,
+      "loss": 0.086,
+      "step": 12404
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 8.359901900674433e-06,
+      "loss": 0.151,
+      "step": 12405
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 8.359768613547304e-06,
+      "loss": 0.0852,
+      "step": 12406
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 8.359635326420175e-06,
+      "loss": 0.1455,
+      "step": 12407
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 8.359502039293046e-06,
+      "loss": 0.1339,
+      "step": 12408
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 8.359368752165917e-06,
+      "loss": 0.1186,
+      "step": 12409
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 8.359235465038788e-06,
+      "loss": 0.1264,
+      "step": 12410
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 8.359102177911657e-06,
+      "loss": 0.1265,
+      "step": 12411
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 8.358968890784528e-06,
+      "loss": 0.1262,
+      "step": 12412
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 8.3588356036574e-06,
+      "loss": 0.0771,
+      "step": 12413
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 8.35870231653027e-06,
+      "loss": 0.0901,
+      "step": 12414
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 8.358569029403141e-06,
+      "loss": 0.0876,
+      "step": 12415
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 8.358435742276012e-06,
+      "loss": 0.1168,
+      "step": 12416
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 8.358302455148881e-06,
+      "loss": 0.1195,
+      "step": 12417
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 8.358169168021754e-06,
+      "loss": 0.0801,
+      "step": 12418
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 8.358035880894623e-06,
+      "loss": 0.1136,
+      "step": 12419
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 8.357902593767494e-06,
+      "loss": 0.1168,
+      "step": 12420
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 8.357769306640365e-06,
+      "loss": 0.0767,
+      "step": 12421
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 8.357636019513235e-06,
+      "loss": 0.0702,
+      "step": 12422
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 8.357502732386107e-06,
+      "loss": 0.0931,
+      "step": 12423
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 8.357369445258977e-06,
+      "loss": 0.1033,
+      "step": 12424
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 8.357236158131848e-06,
+      "loss": 0.1068,
+      "step": 12425
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 8.357102871004719e-06,
+      "loss": 0.1104,
+      "step": 12426
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 8.35696958387759e-06,
+      "loss": 0.083,
+      "step": 12427
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 8.35683629675046e-06,
+      "loss": 0.0773,
+      "step": 12428
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 8.356703009623332e-06,
+      "loss": 0.0922,
+      "step": 12429
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 8.356569722496201e-06,
+      "loss": 0.0874,
+      "step": 12430
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 8.356436435369074e-06,
+      "loss": 0.1144,
+      "step": 12431
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 8.356303148241943e-06,
+      "loss": 0.1323,
+      "step": 12432
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 8.356169861114814e-06,
+      "loss": 0.1237,
+      "step": 12433
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 8.356036573987685e-06,
+      "loss": 0.0983,
+      "step": 12434
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 8.355903286860556e-06,
+      "loss": 0.0967,
+      "step": 12435
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 8.355769999733427e-06,
+      "loss": 0.1192,
+      "step": 12436
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 8.355636712606298e-06,
+      "loss": 0.1278,
+      "step": 12437
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 8.355503425479167e-06,
+      "loss": 0.1194,
+      "step": 12438
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 8.355370138352038e-06,
+      "loss": 0.0808,
+      "step": 12439
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 8.35523685122491e-06,
+      "loss": 0.1215,
+      "step": 12440
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 8.35510356409778e-06,
+      "loss": 0.1187,
+      "step": 12441
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 8.354970276970651e-06,
+      "loss": 0.0872,
+      "step": 12442
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 8.35483698984352e-06,
+      "loss": 0.136,
+      "step": 12443
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 8.354703702716392e-06,
+      "loss": 0.0698,
+      "step": 12444
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 8.354570415589263e-06,
+      "loss": 0.1096,
+      "step": 12445
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 8.354437128462134e-06,
+      "loss": 0.0918,
+      "step": 12446
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 8.354303841335005e-06,
+      "loss": 0.1007,
+      "step": 12447
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 8.354170554207876e-06,
+      "loss": 0.0887,
+      "step": 12448
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 8.354037267080745e-06,
+      "loss": 0.0763,
+      "step": 12449
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 8.353903979953618e-06,
+      "loss": 0.1261,
+      "step": 12450
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 8.353770692826487e-06,
+      "loss": 0.0781,
+      "step": 12451
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 8.353637405699358e-06,
+      "loss": 0.0809,
+      "step": 12452
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 8.353504118572229e-06,
+      "loss": 0.1185,
+      "step": 12453
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 8.3533708314451e-06,
+      "loss": 0.0862,
+      "step": 12454
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 8.35323754431797e-06,
+      "loss": 0.0838,
+      "step": 12455
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 8.353104257190842e-06,
+      "loss": 0.1219,
+      "step": 12456
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 8.352970970063711e-06,
+      "loss": 0.1012,
+      "step": 12457
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 8.352837682936584e-06,
+      "loss": 0.0707,
+      "step": 12458
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 8.352704395809453e-06,
+      "loss": 0.121,
+      "step": 12459
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 8.352571108682324e-06,
+      "loss": 0.0841,
+      "step": 12460
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 8.352437821555195e-06,
+      "loss": 0.1241,
+      "step": 12461
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 8.352304534428064e-06,
+      "loss": 0.1119,
+      "step": 12462
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 8.352171247300937e-06,
+      "loss": 0.0859,
+      "step": 12463
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 8.352037960173806e-06,
+      "loss": 0.0931,
+      "step": 12464
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 8.351904673046677e-06,
+      "loss": 0.0958,
+      "step": 12465
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 8.351771385919548e-06,
+      "loss": 0.113,
+      "step": 12466
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 8.35163809879242e-06,
+      "loss": 0.1366,
+      "step": 12467
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 8.35150481166529e-06,
+      "loss": 0.1203,
+      "step": 12468
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 8.351371524538161e-06,
+      "loss": 0.069,
+      "step": 12469
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 8.35123823741103e-06,
+      "loss": 0.1483,
+      "step": 12470
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 8.351104950283903e-06,
+      "loss": 0.1428,
+      "step": 12471
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 8.350971663156773e-06,
+      "loss": 0.1327,
+      "step": 12472
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 8.350838376029644e-06,
+      "loss": 0.0956,
+      "step": 12473
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 8.350705088902515e-06,
+      "loss": 0.0807,
+      "step": 12474
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 8.350571801775386e-06,
+      "loss": 0.1232,
+      "step": 12475
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 8.350438514648257e-06,
+      "loss": 0.0974,
+      "step": 12476
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 8.350305227521128e-06,
+      "loss": 0.0988,
+      "step": 12477
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 8.350171940393997e-06,
+      "loss": 0.1114,
+      "step": 12478
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 8.350038653266868e-06,
+      "loss": 0.1073,
+      "step": 12479
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 8.349905366139739e-06,
+      "loss": 0.1595,
+      "step": 12480
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 8.34977207901261e-06,
+      "loss": 0.0973,
+      "step": 12481
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 8.349638791885481e-06,
+      "loss": 0.1355,
+      "step": 12482
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 8.34950550475835e-06,
+      "loss": 0.1097,
+      "step": 12483
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 8.349372217631221e-06,
+      "loss": 0.0927,
+      "step": 12484
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 8.349238930504092e-06,
+      "loss": 0.0668,
+      "step": 12485
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 8.349105643376963e-06,
+      "loss": 0.1163,
+      "step": 12486
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 8.348972356249834e-06,
+      "loss": 0.0949,
+      "step": 12487
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 8.348839069122705e-06,
+      "loss": 0.1149,
+      "step": 12488
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 8.348705781995574e-06,
+      "loss": 0.0874,
+      "step": 12489
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 8.348572494868447e-06,
+      "loss": 0.1177,
+      "step": 12490
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 8.348439207741316e-06,
+      "loss": 0.1053,
+      "step": 12491
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 8.348305920614187e-06,
+      "loss": 0.0986,
+      "step": 12492
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 8.348172633487058e-06,
+      "loss": 0.1339,
+      "step": 12493
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 8.34803934635993e-06,
+      "loss": 0.0711,
+      "step": 12494
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 8.3479060592328e-06,
+      "loss": 0.0644,
+      "step": 12495
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 8.347772772105671e-06,
+      "loss": 0.1294,
+      "step": 12496
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 8.34763948497854e-06,
+      "loss": 0.0657,
+      "step": 12497
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 8.347506197851413e-06,
+      "loss": 0.0781,
+      "step": 12498
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 8.347372910724283e-06,
+      "loss": 0.076,
+      "step": 12499
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 8.347239623597154e-06,
+      "loss": 0.0764,
+      "step": 12500
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 8.347106336470025e-06,
+      "loss": 0.1076,
+      "step": 12501
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 8.346973049342894e-06,
+      "loss": 0.0964,
+      "step": 12502
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 8.346839762215767e-06,
+      "loss": 0.099,
+      "step": 12503
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 8.346706475088636e-06,
+      "loss": 0.1036,
+      "step": 12504
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 8.346573187961507e-06,
+      "loss": 0.0989,
+      "step": 12505
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 8.346439900834378e-06,
+      "loss": 0.089,
+      "step": 12506
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 8.346306613707249e-06,
+      "loss": 0.0903,
+      "step": 12507
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 8.34617332658012e-06,
+      "loss": 0.1,
+      "step": 12508
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 8.346040039452991e-06,
+      "loss": 0.114,
+      "step": 12509
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 8.34590675232586e-06,
+      "loss": 0.1193,
+      "step": 12510
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 8.345773465198731e-06,
+      "loss": 0.1293,
+      "step": 12511
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 8.345640178071602e-06,
+      "loss": 0.0887,
+      "step": 12512
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 8.345506890944473e-06,
+      "loss": 0.0821,
+      "step": 12513
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 8.345373603817344e-06,
+      "loss": 0.0776,
+      "step": 12514
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 8.345240316690215e-06,
+      "loss": 0.0949,
+      "step": 12515
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 8.345107029563085e-06,
+      "loss": 0.1079,
+      "step": 12516
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 8.344973742435957e-06,
+      "loss": 0.1313,
+      "step": 12517
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 8.344840455308827e-06,
+      "loss": 0.0679,
+      "step": 12518
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 8.344707168181698e-06,
+      "loss": 0.0814,
+      "step": 12519
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 8.344573881054569e-06,
+      "loss": 0.1123,
+      "step": 12520
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 8.344440593927438e-06,
+      "loss": 0.1433,
+      "step": 12521
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 8.34430730680031e-06,
+      "loss": 0.1003,
+      "step": 12522
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 8.34417401967318e-06,
+      "loss": 0.0879,
+      "step": 12523
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 8.34404073254605e-06,
+      "loss": 0.1306,
+      "step": 12524
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 8.343907445418922e-06,
+      "loss": 0.1042,
+      "step": 12525
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 8.343774158291793e-06,
+      "loss": 0.1198,
+      "step": 12526
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 8.343640871164664e-06,
+      "loss": 0.0932,
+      "step": 12527
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 8.343507584037535e-06,
+      "loss": 0.1023,
+      "step": 12528
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 8.343374296910404e-06,
+      "loss": 0.1011,
+      "step": 12529
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 8.343241009783277e-06,
+      "loss": 0.078,
+      "step": 12530
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 8.343107722656146e-06,
+      "loss": 0.135,
+      "step": 12531
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 8.342974435529017e-06,
+      "loss": 0.1078,
+      "step": 12532
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 8.342841148401888e-06,
+      "loss": 0.0831,
+      "step": 12533
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 8.342707861274759e-06,
+      "loss": 0.0699,
+      "step": 12534
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 8.34257457414763e-06,
+      "loss": 0.1037,
+      "step": 12535
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 8.342441287020501e-06,
+      "loss": 0.0664,
+      "step": 12536
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 8.34230799989337e-06,
+      "loss": 0.0947,
+      "step": 12537
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 8.342174712766243e-06,
+      "loss": 0.1068,
+      "step": 12538
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 8.342041425639112e-06,
+      "loss": 0.0736,
+      "step": 12539
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 8.341908138511983e-06,
+      "loss": 0.1571,
+      "step": 12540
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 8.341774851384854e-06,
+      "loss": 0.1098,
+      "step": 12541
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 8.341641564257724e-06,
+      "loss": 0.1103,
+      "step": 12542
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 8.341508277130596e-06,
+      "loss": 0.099,
+      "step": 12543
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 8.341374990003466e-06,
+      "loss": 0.0659,
+      "step": 12544
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 8.341241702876337e-06,
+      "loss": 0.0879,
+      "step": 12545
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 8.341108415749208e-06,
+      "loss": 0.1368,
+      "step": 12546
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 8.340975128622079e-06,
+      "loss": 0.1494,
+      "step": 12547
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 8.34084184149495e-06,
+      "loss": 0.0852,
+      "step": 12548
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 8.34070855436782e-06,
+      "loss": 0.1261,
+      "step": 12549
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 8.34057526724069e-06,
+      "loss": 0.0989,
+      "step": 12550
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 8.340441980113561e-06,
+      "loss": 0.0858,
+      "step": 12551
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 8.340308692986432e-06,
+      "loss": 0.0834,
+      "step": 12552
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 8.340175405859303e-06,
+      "loss": 0.0914,
+      "step": 12553
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 8.340042118732174e-06,
+      "loss": 0.119,
+      "step": 12554
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 8.339908831605045e-06,
+      "loss": 0.132,
+      "step": 12555
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 8.339775544477914e-06,
+      "loss": 0.1217,
+      "step": 12556
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 8.339642257350787e-06,
+      "loss": 0.1222,
+      "step": 12557
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 8.339508970223656e-06,
+      "loss": 0.0766,
+      "step": 12558
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 8.339375683096527e-06,
+      "loss": 0.081,
+      "step": 12559
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 8.339242395969398e-06,
+      "loss": 0.1203,
+      "step": 12560
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 8.339109108842267e-06,
+      "loss": 0.0855,
+      "step": 12561
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 8.33897582171514e-06,
+      "loss": 0.0826,
+      "step": 12562
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 8.33884253458801e-06,
+      "loss": 0.0816,
+      "step": 12563
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 8.33870924746088e-06,
+      "loss": 0.0731,
+      "step": 12564
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 8.338575960333751e-06,
+      "loss": 0.1032,
+      "step": 12565
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 8.338442673206622e-06,
+      "loss": 0.1053,
+      "step": 12566
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 8.338309386079493e-06,
+      "loss": 0.0588,
+      "step": 12567
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 8.338176098952364e-06,
+      "loss": 0.0947,
+      "step": 12568
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 8.338042811825234e-06,
+      "loss": 0.0752,
+      "step": 12569
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 8.337909524698106e-06,
+      "loss": 0.1045,
+      "step": 12570
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 8.337776237570976e-06,
+      "loss": 0.1251,
+      "step": 12571
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 8.337642950443847e-06,
+      "loss": 0.0855,
+      "step": 12572
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 8.337509663316718e-06,
+      "loss": 0.0676,
+      "step": 12573
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 8.337376376189589e-06,
+      "loss": 0.0902,
+      "step": 12574
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 8.33724308906246e-06,
+      "loss": 0.1181,
+      "step": 12575
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 8.33710980193533e-06,
+      "loss": 0.1328,
+      "step": 12576
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 8.3369765148082e-06,
+      "loss": 0.0905,
+      "step": 12577
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 8.336843227681071e-06,
+      "loss": 0.1105,
+      "step": 12578
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 8.336709940553942e-06,
+      "loss": 0.1205,
+      "step": 12579
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 8.336576653426813e-06,
+      "loss": 0.0969,
+      "step": 12580
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 8.336443366299684e-06,
+      "loss": 0.0858,
+      "step": 12581
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 8.336310079172553e-06,
+      "loss": 0.0998,
+      "step": 12582
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 8.336176792045424e-06,
+      "loss": 0.1357,
+      "step": 12583
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 8.336043504918295e-06,
+      "loss": 0.0784,
+      "step": 12584
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 8.335910217791166e-06,
+      "loss": 0.0736,
+      "step": 12585
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 8.335776930664037e-06,
+      "loss": 0.0919,
+      "step": 12586
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 8.335643643536908e-06,
+      "loss": 0.0798,
+      "step": 12587
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 8.335510356409778e-06,
+      "loss": 0.1213,
+      "step": 12588
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 8.33537706928265e-06,
+      "loss": 0.0969,
+      "step": 12589
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 8.33524378215552e-06,
+      "loss": 0.0811,
+      "step": 12590
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 8.33511049502839e-06,
+      "loss": 0.1239,
+      "step": 12591
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 8.334977207901262e-06,
+      "loss": 0.1053,
+      "step": 12592
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 8.334843920774133e-06,
+      "loss": 0.1079,
+      "step": 12593
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 8.334710633647004e-06,
+      "loss": 0.0853,
+      "step": 12594
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 8.334577346519875e-06,
+      "loss": 0.1149,
+      "step": 12595
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 8.334444059392744e-06,
+      "loss": 0.0626,
+      "step": 12596
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 8.334310772265616e-06,
+      "loss": 0.1459,
+      "step": 12597
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 8.334177485138486e-06,
+      "loss": 0.0981,
+      "step": 12598
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 8.334044198011357e-06,
+      "loss": 0.0946,
+      "step": 12599
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 8.333910910884228e-06,
+      "loss": 0.0914,
+      "step": 12600
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 8.333777623757097e-06,
+      "loss": 0.1022,
+      "step": 12601
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 8.33364433662997e-06,
+      "loss": 0.1127,
+      "step": 12602
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 8.333511049502839e-06,
+      "loss": 0.0889,
+      "step": 12603
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 8.33337776237571e-06,
+      "loss": 0.0916,
+      "step": 12604
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 8.333244475248581e-06,
+      "loss": 0.1378,
+      "step": 12605
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 8.333111188121452e-06,
+      "loss": 0.0902,
+      "step": 12606
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 8.332977900994323e-06,
+      "loss": 0.093,
+      "step": 12607
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 8.332844613867194e-06,
+      "loss": 0.0962,
+      "step": 12608
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 8.332711326740063e-06,
+      "loss": 0.1306,
+      "step": 12609
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 8.332578039612936e-06,
+      "loss": 0.1436,
+      "step": 12610
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 8.332444752485805e-06,
+      "loss": 0.0625,
+      "step": 12611
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 8.332311465358676e-06,
+      "loss": 0.0839,
+      "step": 12612
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 8.332178178231547e-06,
+      "loss": 0.0833,
+      "step": 12613
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 8.332044891104418e-06,
+      "loss": 0.0901,
+      "step": 12614
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 8.33191160397729e-06,
+      "loss": 0.1127,
+      "step": 12615
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 8.33177831685016e-06,
+      "loss": 0.1141,
+      "step": 12616
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 8.33164502972303e-06,
+      "loss": 0.0958,
+      "step": 12617
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 8.3315117425959e-06,
+      "loss": 0.1056,
+      "step": 12618
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 8.331378455468772e-06,
+      "loss": 0.0931,
+      "step": 12619
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 8.331245168341643e-06,
+      "loss": 0.1119,
+      "step": 12620
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 8.331111881214514e-06,
+      "loss": 0.1122,
+      "step": 12621
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 8.330978594087383e-06,
+      "loss": 0.1175,
+      "step": 12622
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 8.330845306960254e-06,
+      "loss": 0.1085,
+      "step": 12623
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 8.330712019833125e-06,
+      "loss": 0.1092,
+      "step": 12624
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 8.330578732705996e-06,
+      "loss": 0.0598,
+      "step": 12625
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 8.330445445578867e-06,
+      "loss": 0.0803,
+      "step": 12626
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 8.330312158451738e-06,
+      "loss": 0.0758,
+      "step": 12627
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 8.330178871324607e-06,
+      "loss": 0.1495,
+      "step": 12628
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 8.33004558419748e-06,
+      "loss": 0.1054,
+      "step": 12629
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 8.32991229707035e-06,
+      "loss": 0.1151,
+      "step": 12630
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 8.32977900994322e-06,
+      "loss": 0.126,
+      "step": 12631
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 8.329645722816091e-06,
+      "loss": 0.1194,
+      "step": 12632
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 8.329512435688962e-06,
+      "loss": 0.088,
+      "step": 12633
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 8.329379148561833e-06,
+      "loss": 0.1028,
+      "step": 12634
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 8.329245861434704e-06,
+      "loss": 0.0688,
+      "step": 12635
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 8.329112574307573e-06,
+      "loss": 0.0821,
+      "step": 12636
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 8.328979287180444e-06,
+      "loss": 0.093,
+      "step": 12637
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 8.328846000053315e-06,
+      "loss": 0.1141,
+      "step": 12638
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 8.328712712926186e-06,
+      "loss": 0.086,
+      "step": 12639
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 8.328579425799057e-06,
+      "loss": 0.0906,
+      "step": 12640
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 8.328446138671927e-06,
+      "loss": 0.085,
+      "step": 12641
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 8.3283128515448e-06,
+      "loss": 0.1244,
+      "step": 12642
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 8.328179564417669e-06,
+      "loss": 0.0777,
+      "step": 12643
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 8.32804627729054e-06,
+      "loss": 0.1122,
+      "step": 12644
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 8.32791299016341e-06,
+      "loss": 0.0986,
+      "step": 12645
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 8.327779703036282e-06,
+      "loss": 0.1097,
+      "step": 12646
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 8.327646415909153e-06,
+      "loss": 0.1116,
+      "step": 12647
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 8.327513128782024e-06,
+      "loss": 0.1139,
+      "step": 12648
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 8.327379841654893e-06,
+      "loss": 0.1172,
+      "step": 12649
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 8.327246554527764e-06,
+      "loss": 0.1792,
+      "step": 12650
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 8.327113267400635e-06,
+      "loss": 0.1482,
+      "step": 12651
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 8.326979980273506e-06,
+      "loss": 0.0878,
+      "step": 12652
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 8.326846693146377e-06,
+      "loss": 0.0709,
+      "step": 12653
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 8.326713406019248e-06,
+      "loss": 0.1196,
+      "step": 12654
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 8.326580118892117e-06,
+      "loss": 0.114,
+      "step": 12655
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 8.32644683176499e-06,
+      "loss": 0.1079,
+      "step": 12656
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 8.32631354463786e-06,
+      "loss": 0.1284,
+      "step": 12657
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 8.32618025751073e-06,
+      "loss": 0.0926,
+      "step": 12658
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 8.326046970383601e-06,
+      "loss": 0.0855,
+      "step": 12659
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 8.32591368325647e-06,
+      "loss": 0.1412,
+      "step": 12660
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 8.325780396129343e-06,
+      "loss": 0.0684,
+      "step": 12661
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 8.325647109002213e-06,
+      "loss": 0.0902,
+      "step": 12662
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 8.325513821875084e-06,
+      "loss": 0.078,
+      "step": 12663
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 8.325380534747955e-06,
+      "loss": 0.1363,
+      "step": 12664
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 8.325247247620825e-06,
+      "loss": 0.1773,
+      "step": 12665
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 8.325113960493696e-06,
+      "loss": 0.1047,
+      "step": 12666
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 8.324980673366567e-06,
+      "loss": 0.1045,
+      "step": 12667
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 8.324847386239437e-06,
+      "loss": 0.0926,
+      "step": 12668
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 8.32471409911231e-06,
+      "loss": 0.1105,
+      "step": 12669
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 8.324580811985179e-06,
+      "loss": 0.1164,
+      "step": 12670
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 8.32444752485805e-06,
+      "loss": 0.0866,
+      "step": 12671
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 8.32431423773092e-06,
+      "loss": 0.0764,
+      "step": 12672
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 8.324180950603792e-06,
+      "loss": 0.1137,
+      "step": 12673
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 8.324047663476663e-06,
+      "loss": 0.0992,
+      "step": 12674
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 8.323914376349534e-06,
+      "loss": 0.1034,
+      "step": 12675
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 8.323781089222403e-06,
+      "loss": 0.1468,
+      "step": 12676
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 8.323647802095274e-06,
+      "loss": 0.0843,
+      "step": 12677
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 8.323514514968145e-06,
+      "loss": 0.0882,
+      "step": 12678
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 8.323381227841016e-06,
+      "loss": 0.1081,
+      "step": 12679
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 8.323247940713887e-06,
+      "loss": 0.1027,
+      "step": 12680
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 8.323114653586756e-06,
+      "loss": 0.0876,
+      "step": 12681
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 8.322981366459629e-06,
+      "loss": 0.0822,
+      "step": 12682
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 8.322848079332498e-06,
+      "loss": 0.0787,
+      "step": 12683
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 8.32271479220537e-06,
+      "loss": 0.1076,
+      "step": 12684
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 8.32258150507824e-06,
+      "loss": 0.1149,
+      "step": 12685
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 8.322448217951111e-06,
+      "loss": 0.0865,
+      "step": 12686
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 8.322314930823982e-06,
+      "loss": 0.1022,
+      "step": 12687
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 8.322181643696853e-06,
+      "loss": 0.0919,
+      "step": 12688
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 8.322048356569723e-06,
+      "loss": 0.0756,
+      "step": 12689
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 8.321915069442594e-06,
+      "loss": 0.08,
+      "step": 12690
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 8.321781782315465e-06,
+      "loss": 0.0807,
+      "step": 12691
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 8.321648495188336e-06,
+      "loss": 0.0983,
+      "step": 12692
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 8.321515208061207e-06,
+      "loss": 0.1091,
+      "step": 12693
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 8.321381920934078e-06,
+      "loss": 0.0875,
+      "step": 12694
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 8.321248633806947e-06,
+      "loss": 0.1726,
+      "step": 12695
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 8.321115346679818e-06,
+      "loss": 0.1573,
+      "step": 12696
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 8.320982059552689e-06,
+      "loss": 0.1107,
+      "step": 12697
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 8.32084877242556e-06,
+      "loss": 0.1125,
+      "step": 12698
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 8.320715485298431e-06,
+      "loss": 0.0775,
+      "step": 12699
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 8.3205821981713e-06,
+      "loss": 0.1166,
+      "step": 12700
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 8.320448911044173e-06,
+      "loss": 0.0719,
+      "step": 12701
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 8.320315623917042e-06,
+      "loss": 0.1018,
+      "step": 12702
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 8.320182336789913e-06,
+      "loss": 0.1258,
+      "step": 12703
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 8.320049049662784e-06,
+      "loss": 0.1258,
+      "step": 12704
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 8.319915762535655e-06,
+      "loss": 0.0841,
+      "step": 12705
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 8.319782475408526e-06,
+      "loss": 0.1149,
+      "step": 12706
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 8.319649188281397e-06,
+      "loss": 0.0671,
+      "step": 12707
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 8.319515901154266e-06,
+      "loss": 0.1189,
+      "step": 12708
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 8.319382614027139e-06,
+      "loss": 0.1325,
+      "step": 12709
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 8.319249326900008e-06,
+      "loss": 0.115,
+      "step": 12710
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 8.31911603977288e-06,
+      "loss": 0.0868,
+      "step": 12711
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 8.31898275264575e-06,
+      "loss": 0.1057,
+      "step": 12712
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 8.318849465518621e-06,
+      "loss": 0.0797,
+      "step": 12713
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 8.318716178391492e-06,
+      "loss": 0.1153,
+      "step": 12714
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 8.318582891264363e-06,
+      "loss": 0.1282,
+      "step": 12715
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 8.318449604137233e-06,
+      "loss": 0.065,
+      "step": 12716
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 8.318316317010104e-06,
+      "loss": 0.0775,
+      "step": 12717
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 8.318183029882975e-06,
+      "loss": 0.1079,
+      "step": 12718
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 8.318049742755846e-06,
+      "loss": 0.1056,
+      "step": 12719
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 8.317916455628717e-06,
+      "loss": 0.0859,
+      "step": 12720
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 8.317783168501586e-06,
+      "loss": 0.0976,
+      "step": 12721
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 8.317649881374457e-06,
+      "loss": 0.0995,
+      "step": 12722
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 8.317516594247328e-06,
+      "loss": 0.1148,
+      "step": 12723
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 8.317383307120199e-06,
+      "loss": 0.1109,
+      "step": 12724
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 8.31725001999307e-06,
+      "loss": 0.1116,
+      "step": 12725
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 8.317116732865941e-06,
+      "loss": 0.1171,
+      "step": 12726
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 8.31698344573881e-06,
+      "loss": 0.1187,
+      "step": 12727
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 8.316850158611683e-06,
+      "loss": 0.1171,
+      "step": 12728
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 8.316716871484552e-06,
+      "loss": 0.0796,
+      "step": 12729
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 8.316583584357423e-06,
+      "loss": 0.0765,
+      "step": 12730
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 8.316450297230294e-06,
+      "loss": 0.1159,
+      "step": 12731
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 8.316317010103165e-06,
+      "loss": 0.1392,
+      "step": 12732
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 8.316183722976036e-06,
+      "loss": 0.1281,
+      "step": 12733
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 8.316050435848907e-06,
+      "loss": 0.1356,
+      "step": 12734
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 8.315917148721776e-06,
+      "loss": 0.0868,
+      "step": 12735
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 8.315783861594647e-06,
+      "loss": 0.1094,
+      "step": 12736
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 8.315650574467518e-06,
+      "loss": 0.119,
+      "step": 12737
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 8.31551728734039e-06,
+      "loss": 0.0993,
+      "step": 12738
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 8.31538400021326e-06,
+      "loss": 0.0708,
+      "step": 12739
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 8.31525071308613e-06,
+      "loss": 0.1015,
+      "step": 12740
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 8.315117425959002e-06,
+      "loss": 0.1137,
+      "step": 12741
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 8.314984138831872e-06,
+      "loss": 0.1575,
+      "step": 12742
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 8.314850851704743e-06,
+      "loss": 0.1264,
+      "step": 12743
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 8.314717564577614e-06,
+      "loss": 0.0732,
+      "step": 12744
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 8.314584277450485e-06,
+      "loss": 0.2137,
+      "step": 12745
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 8.314450990323356e-06,
+      "loss": 0.1271,
+      "step": 12746
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 8.314317703196227e-06,
+      "loss": 0.0865,
+      "step": 12747
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 8.314184416069096e-06,
+      "loss": 0.0667,
+      "step": 12748
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 8.314051128941967e-06,
+      "loss": 0.0941,
+      "step": 12749
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 8.313917841814838e-06,
+      "loss": 0.1335,
+      "step": 12750
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 8.313784554687709e-06,
+      "loss": 0.1121,
+      "step": 12751
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 8.31365126756058e-06,
+      "loss": 0.1142,
+      "step": 12752
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 8.313517980433451e-06,
+      "loss": 0.0911,
+      "step": 12753
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 8.31338469330632e-06,
+      "loss": 0.109,
+      "step": 12754
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 8.313251406179191e-06,
+      "loss": 0.1013,
+      "step": 12755
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 8.313118119052062e-06,
+      "loss": 0.0977,
+      "step": 12756
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 8.312984831924933e-06,
+      "loss": 0.109,
+      "step": 12757
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 8.312851544797804e-06,
+      "loss": 0.1159,
+      "step": 12758
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 8.312718257670674e-06,
+      "loss": 0.1069,
+      "step": 12759
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 8.312584970543546e-06,
+      "loss": 0.0871,
+      "step": 12760
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 8.312451683416416e-06,
+      "loss": 0.0962,
+      "step": 12761
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 8.312318396289287e-06,
+      "loss": 0.1386,
+      "step": 12762
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 8.312185109162158e-06,
+      "loss": 0.0643,
+      "step": 12763
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 8.312051822035029e-06,
+      "loss": 0.1024,
+      "step": 12764
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 8.3119185349079e-06,
+      "loss": 0.1144,
+      "step": 12765
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 8.31178524778077e-06,
+      "loss": 0.0786,
+      "step": 12766
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 8.31165196065364e-06,
+      "loss": 0.1198,
+      "step": 12767
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 8.311518673526513e-06,
+      "loss": 0.1202,
+      "step": 12768
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 8.311385386399382e-06,
+      "loss": 0.1028,
+      "step": 12769
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 8.311252099272253e-06,
+      "loss": 0.0839,
+      "step": 12770
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 8.311118812145124e-06,
+      "loss": 0.083,
+      "step": 12771
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 8.310985525017995e-06,
+      "loss": 0.1194,
+      "step": 12772
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 8.310852237890866e-06,
+      "loss": 0.0839,
+      "step": 12773
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 8.310718950763737e-06,
+      "loss": 0.1072,
+      "step": 12774
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 8.310585663636606e-06,
+      "loss": 0.115,
+      "step": 12775
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 8.310452376509477e-06,
+      "loss": 0.119,
+      "step": 12776
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 8.310319089382348e-06,
+      "loss": 0.1427,
+      "step": 12777
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 8.310185802255219e-06,
+      "loss": 0.0879,
+      "step": 12778
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 8.31005251512809e-06,
+      "loss": 0.0695,
+      "step": 12779
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 8.30991922800096e-06,
+      "loss": 0.0975,
+      "step": 12780
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 8.309785940873832e-06,
+      "loss": 0.0773,
+      "step": 12781
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 8.309652653746701e-06,
+      "loss": 0.1092,
+      "step": 12782
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 8.309519366619572e-06,
+      "loss": 0.0951,
+      "step": 12783
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 8.309386079492443e-06,
+      "loss": 0.0835,
+      "step": 12784
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 8.309252792365314e-06,
+      "loss": 0.1009,
+      "step": 12785
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 8.309119505238185e-06,
+      "loss": 0.0922,
+      "step": 12786
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 8.308986218111056e-06,
+      "loss": 0.1026,
+      "step": 12787
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 8.308852930983926e-06,
+      "loss": 0.0803,
+      "step": 12788
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 8.308719643856797e-06,
+      "loss": 0.0995,
+      "step": 12789
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 8.308586356729668e-06,
+      "loss": 0.0795,
+      "step": 12790
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 8.308453069602539e-06,
+      "loss": 0.0914,
+      "step": 12791
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 8.30831978247541e-06,
+      "loss": 0.1069,
+      "step": 12792
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 8.30818649534828e-06,
+      "loss": 0.1531,
+      "step": 12793
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 8.30805320822115e-06,
+      "loss": 0.0691,
+      "step": 12794
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 8.307919921094021e-06,
+      "loss": 0.0667,
+      "step": 12795
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 8.307786633966892e-06,
+      "loss": 0.101,
+      "step": 12796
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 8.307653346839763e-06,
+      "loss": 0.1031,
+      "step": 12797
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 8.307520059712634e-06,
+      "loss": 0.0738,
+      "step": 12798
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 8.307386772585503e-06,
+      "loss": 0.1595,
+      "step": 12799
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 8.307253485458376e-06,
+      "loss": 0.0774,
+      "step": 12800
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 8.307120198331245e-06,
+      "loss": 0.1132,
+      "step": 12801
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 8.306986911204116e-06,
+      "loss": 0.1314,
+      "step": 12802
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 8.306853624076987e-06,
+      "loss": 0.1008,
+      "step": 12803
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 8.306720336949858e-06,
+      "loss": 0.091,
+      "step": 12804
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 8.30658704982273e-06,
+      "loss": 0.0952,
+      "step": 12805
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 8.3064537626956e-06,
+      "loss": 0.1381,
+      "step": 12806
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 8.30632047556847e-06,
+      "loss": 0.1586,
+      "step": 12807
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 8.306187188441342e-06,
+      "loss": 0.1248,
+      "step": 12808
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 8.306053901314211e-06,
+      "loss": 0.0924,
+      "step": 12809
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 8.305920614187082e-06,
+      "loss": 0.0935,
+      "step": 12810
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 8.305787327059953e-06,
+      "loss": 0.0931,
+      "step": 12811
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 8.305654039932824e-06,
+      "loss": 0.0801,
+      "step": 12812
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 8.305520752805695e-06,
+      "loss": 0.1175,
+      "step": 12813
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 8.305387465678565e-06,
+      "loss": 0.1059,
+      "step": 12814
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 8.305254178551436e-06,
+      "loss": 0.1262,
+      "step": 12815
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 8.305120891424307e-06,
+      "loss": 0.1138,
+      "step": 12816
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 8.304987604297178e-06,
+      "loss": 0.1382,
+      "step": 12817
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 8.304854317170049e-06,
+      "loss": 0.0845,
+      "step": 12818
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 8.30472103004292e-06,
+      "loss": 0.0814,
+      "step": 12819
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 8.304587742915789e-06,
+      "loss": 0.0738,
+      "step": 12820
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 8.30445445578866e-06,
+      "loss": 0.1276,
+      "step": 12821
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 8.304321168661531e-06,
+      "loss": 0.0793,
+      "step": 12822
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 8.304187881534402e-06,
+      "loss": 0.1411,
+      "step": 12823
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 8.304054594407273e-06,
+      "loss": 0.0806,
+      "step": 12824
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 8.303921307280144e-06,
+      "loss": 0.1338,
+      "step": 12825
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 8.303788020153013e-06,
+      "loss": 0.1209,
+      "step": 12826
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 8.303654733025886e-06,
+      "loss": 0.084,
+      "step": 12827
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 8.303521445898755e-06,
+      "loss": 0.1604,
+      "step": 12828
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 8.303388158771626e-06,
+      "loss": 0.1127,
+      "step": 12829
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 8.303254871644497e-06,
+      "loss": 0.13,
+      "step": 12830
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 8.303121584517368e-06,
+      "loss": 0.1052,
+      "step": 12831
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 8.30298829739024e-06,
+      "loss": 0.087,
+      "step": 12832
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 8.302855010263109e-06,
+      "loss": 0.1322,
+      "step": 12833
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 8.30272172313598e-06,
+      "loss": 0.151,
+      "step": 12834
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 8.30258843600885e-06,
+      "loss": 0.1234,
+      "step": 12835
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 8.302455148881722e-06,
+      "loss": 0.1392,
+      "step": 12836
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 8.302321861754593e-06,
+      "loss": 0.0841,
+      "step": 12837
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 8.302188574627464e-06,
+      "loss": 0.1287,
+      "step": 12838
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 8.302055287500333e-06,
+      "loss": 0.116,
+      "step": 12839
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 8.301922000373206e-06,
+      "loss": 0.0821,
+      "step": 12840
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 8.301788713246075e-06,
+      "loss": 0.1117,
+      "step": 12841
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 8.301655426118946e-06,
+      "loss": 0.0891,
+      "step": 12842
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 8.301522138991817e-06,
+      "loss": 0.1156,
+      "step": 12843
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 8.301388851864688e-06,
+      "loss": 0.0722,
+      "step": 12844
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 8.301255564737559e-06,
+      "loss": 0.0683,
+      "step": 12845
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 8.30112227761043e-06,
+      "loss": 0.1038,
+      "step": 12846
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 8.300988990483299e-06,
+      "loss": 0.0914,
+      "step": 12847
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 8.300855703356172e-06,
+      "loss": 0.1139,
+      "step": 12848
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 8.300722416229041e-06,
+      "loss": 0.1203,
+      "step": 12849
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 8.300589129101912e-06,
+      "loss": 0.0969,
+      "step": 12850
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 8.300455841974783e-06,
+      "loss": 0.0885,
+      "step": 12851
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 8.300322554847654e-06,
+      "loss": 0.1336,
+      "step": 12852
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 8.300189267720525e-06,
+      "loss": 0.0776,
+      "step": 12853
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 8.300055980593394e-06,
+      "loss": 0.1034,
+      "step": 12854
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 8.299922693466265e-06,
+      "loss": 0.0565,
+      "step": 12855
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 8.299789406339136e-06,
+      "loss": 0.1008,
+      "step": 12856
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 8.299656119212007e-06,
+      "loss": 0.0742,
+      "step": 12857
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 8.299522832084878e-06,
+      "loss": 0.0806,
+      "step": 12858
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 8.29938954495775e-06,
+      "loss": 0.0956,
+      "step": 12859
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 8.299256257830619e-06,
+      "loss": 0.062,
+      "step": 12860
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 8.29912297070349e-06,
+      "loss": 0.1148,
+      "step": 12861
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 8.29898968357636e-06,
+      "loss": 0.1203,
+      "step": 12862
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 8.298856396449232e-06,
+      "loss": 0.0732,
+      "step": 12863
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 8.298723109322103e-06,
+      "loss": 0.1105,
+      "step": 12864
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 8.298589822194974e-06,
+      "loss": 0.1075,
+      "step": 12865
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 8.298456535067843e-06,
+      "loss": 0.102,
+      "step": 12866
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 8.298323247940716e-06,
+      "loss": 0.1095,
+      "step": 12867
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 8.298189960813585e-06,
+      "loss": 0.0622,
+      "step": 12868
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 8.298056673686456e-06,
+      "loss": 0.1354,
+      "step": 12869
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 8.297923386559327e-06,
+      "loss": 0.127,
+      "step": 12870
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 8.297790099432198e-06,
+      "loss": 0.0899,
+      "step": 12871
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 8.297656812305069e-06,
+      "loss": 0.1171,
+      "step": 12872
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 8.297523525177938e-06,
+      "loss": 0.0971,
+      "step": 12873
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 8.29739023805081e-06,
+      "loss": 0.1041,
+      "step": 12874
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 8.29725695092368e-06,
+      "loss": 0.1249,
+      "step": 12875
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 8.297123663796551e-06,
+      "loss": 0.1159,
+      "step": 12876
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 8.296990376669422e-06,
+      "loss": 0.087,
+      "step": 12877
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 8.296857089542293e-06,
+      "loss": 0.0951,
+      "step": 12878
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 8.296723802415162e-06,
+      "loss": 0.1237,
+      "step": 12879
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 8.296590515288035e-06,
+      "loss": 0.0919,
+      "step": 12880
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 8.296457228160904e-06,
+      "loss": 0.0968,
+      "step": 12881
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 8.296323941033775e-06,
+      "loss": 0.0952,
+      "step": 12882
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 8.296190653906646e-06,
+      "loss": 0.0969,
+      "step": 12883
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 8.296057366779517e-06,
+      "loss": 0.1148,
+      "step": 12884
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 8.295924079652388e-06,
+      "loss": 0.089,
+      "step": 12885
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 8.29579079252526e-06,
+      "loss": 0.0786,
+      "step": 12886
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 8.295657505398129e-06,
+      "loss": 0.1031,
+      "step": 12887
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 8.295524218271e-06,
+      "loss": 0.1064,
+      "step": 12888
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 8.29539093114387e-06,
+      "loss": 0.1019,
+      "step": 12889
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 8.295257644016742e-06,
+      "loss": 0.0756,
+      "step": 12890
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 8.295124356889613e-06,
+      "loss": 0.1365,
+      "step": 12891
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 8.294991069762482e-06,
+      "loss": 0.1054,
+      "step": 12892
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 8.294857782635353e-06,
+      "loss": 0.0625,
+      "step": 12893
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 8.294724495508224e-06,
+      "loss": 0.0771,
+      "step": 12894
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 8.294591208381095e-06,
+      "loss": 0.0872,
+      "step": 12895
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 8.294457921253966e-06,
+      "loss": 0.0972,
+      "step": 12896
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 8.294324634126837e-06,
+      "loss": 0.0912,
+      "step": 12897
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 8.294191346999706e-06,
+      "loss": 0.0931,
+      "step": 12898
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 8.294058059872579e-06,
+      "loss": 0.1081,
+      "step": 12899
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 8.293924772745448e-06,
+      "loss": 0.1319,
+      "step": 12900
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 8.29379148561832e-06,
+      "loss": 0.0635,
+      "step": 12901
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 8.29365819849119e-06,
+      "loss": 0.0868,
+      "step": 12902
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 8.293524911364061e-06,
+      "loss": 0.121,
+      "step": 12903
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 8.293391624236932e-06,
+      "loss": 0.0821,
+      "step": 12904
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 8.293258337109803e-06,
+      "loss": 0.1017,
+      "step": 12905
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 8.293125049982673e-06,
+      "loss": 0.0671,
+      "step": 12906
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 8.292991762855545e-06,
+      "loss": 0.0958,
+      "step": 12907
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 8.292858475728415e-06,
+      "loss": 0.0836,
+      "step": 12908
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 8.292725188601286e-06,
+      "loss": 0.121,
+      "step": 12909
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 8.292591901474157e-06,
+      "loss": 0.1094,
+      "step": 12910
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 8.292458614347028e-06,
+      "loss": 0.1332,
+      "step": 12911
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 8.292325327219899e-06,
+      "loss": 0.1008,
+      "step": 12912
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 8.292192040092768e-06,
+      "loss": 0.0937,
+      "step": 12913
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 8.292058752965639e-06,
+      "loss": 0.1093,
+      "step": 12914
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 8.29192546583851e-06,
+      "loss": 0.1252,
+      "step": 12915
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 8.29179217871138e-06,
+      "loss": 0.1149,
+      "step": 12916
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 8.291658891584252e-06,
+      "loss": 0.1038,
+      "step": 12917
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 8.291525604457123e-06,
+      "loss": 0.0759,
+      "step": 12918
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 8.291392317329992e-06,
+      "loss": 0.083,
+      "step": 12919
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 8.291259030202865e-06,
+      "loss": 0.1097,
+      "step": 12920
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 8.291125743075734e-06,
+      "loss": 0.1336,
+      "step": 12921
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 8.290992455948605e-06,
+      "loss": 0.1319,
+      "step": 12922
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 8.290859168821476e-06,
+      "loss": 0.1263,
+      "step": 12923
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 8.290725881694347e-06,
+      "loss": 0.0955,
+      "step": 12924
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 8.290592594567218e-06,
+      "loss": 0.0928,
+      "step": 12925
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 8.290459307440089e-06,
+      "loss": 0.1402,
+      "step": 12926
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 8.290326020312958e-06,
+      "loss": 0.0667,
+      "step": 12927
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 8.29019273318583e-06,
+      "loss": 0.0961,
+      "step": 12928
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 8.2900594460587e-06,
+      "loss": 0.0963,
+      "step": 12929
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 8.289926158931571e-06,
+      "loss": 0.0692,
+      "step": 12930
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 8.289792871804442e-06,
+      "loss": 0.1185,
+      "step": 12931
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 8.289659584677312e-06,
+      "loss": 0.0948,
+      "step": 12932
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 8.289526297550183e-06,
+      "loss": 0.1025,
+      "step": 12933
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 8.289393010423054e-06,
+      "loss": 0.0744,
+      "step": 12934
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 8.289259723295925e-06,
+      "loss": 0.0814,
+      "step": 12935
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 8.289126436168796e-06,
+      "loss": 0.0892,
+      "step": 12936
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 8.288993149041667e-06,
+      "loss": 0.1068,
+      "step": 12937
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 8.288859861914536e-06,
+      "loss": 0.108,
+      "step": 12938
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 8.288726574787409e-06,
+      "loss": 0.1075,
+      "step": 12939
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 8.288593287660278e-06,
+      "loss": 0.0863,
+      "step": 12940
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 8.288460000533149e-06,
+      "loss": 0.0894,
+      "step": 12941
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 8.28832671340602e-06,
+      "loss": 0.0929,
+      "step": 12942
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 8.288193426278891e-06,
+      "loss": 0.1002,
+      "step": 12943
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 8.288060139151762e-06,
+      "loss": 0.1353,
+      "step": 12944
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 8.287926852024633e-06,
+      "loss": 0.0917,
+      "step": 12945
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 8.287793564897502e-06,
+      "loss": 0.1067,
+      "step": 12946
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 8.287660277770375e-06,
+      "loss": 0.0813,
+      "step": 12947
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 8.287526990643244e-06,
+      "loss": 0.0849,
+      "step": 12948
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 8.287393703516115e-06,
+      "loss": 0.1364,
+      "step": 12949
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 8.287260416388986e-06,
+      "loss": 0.1002,
+      "step": 12950
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 8.287127129261855e-06,
+      "loss": 0.094,
+      "step": 12951
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 8.286993842134728e-06,
+      "loss": 0.1142,
+      "step": 12952
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 8.286860555007597e-06,
+      "loss": 0.0782,
+      "step": 12953
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 8.286727267880468e-06,
+      "loss": 0.1172,
+      "step": 12954
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 8.28659398075334e-06,
+      "loss": 0.124,
+      "step": 12955
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 8.28646069362621e-06,
+      "loss": 0.1456,
+      "step": 12956
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 8.286327406499081e-06,
+      "loss": 0.0809,
+      "step": 12957
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 8.286194119371952e-06,
+      "loss": 0.12,
+      "step": 12958
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 8.286060832244822e-06,
+      "loss": 0.1228,
+      "step": 12959
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 8.285927545117693e-06,
+      "loss": 0.1096,
+      "step": 12960
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 8.285794257990564e-06,
+      "loss": 0.1165,
+      "step": 12961
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 8.285660970863435e-06,
+      "loss": 0.1194,
+      "step": 12962
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 8.285527683736306e-06,
+      "loss": 0.1049,
+      "step": 12963
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 8.285394396609177e-06,
+      "loss": 0.0792,
+      "step": 12964
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 8.285261109482046e-06,
+      "loss": 0.0977,
+      "step": 12965
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 8.285127822354919e-06,
+      "loss": 0.1024,
+      "step": 12966
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 8.284994535227788e-06,
+      "loss": 0.1047,
+      "step": 12967
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 8.284861248100659e-06,
+      "loss": 0.1099,
+      "step": 12968
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 8.28472796097353e-06,
+      "loss": 0.0924,
+      "step": 12969
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 8.2845946738464e-06,
+      "loss": 0.1412,
+      "step": 12970
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 8.284461386719272e-06,
+      "loss": 0.1267,
+      "step": 12971
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 8.284328099592141e-06,
+      "loss": 0.0669,
+      "step": 12972
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 8.284194812465012e-06,
+      "loss": 0.0968,
+      "step": 12973
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 8.284061525337883e-06,
+      "loss": 0.1019,
+      "step": 12974
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 8.283928238210754e-06,
+      "loss": 0.0695,
+      "step": 12975
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 8.283794951083625e-06,
+      "loss": 0.0981,
+      "step": 12976
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 8.283661663956496e-06,
+      "loss": 0.0853,
+      "step": 12977
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 8.283528376829366e-06,
+      "loss": 0.0981,
+      "step": 12978
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 8.283395089702238e-06,
+      "loss": 0.0972,
+      "step": 12979
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 8.283261802575108e-06,
+      "loss": 0.0821,
+      "step": 12980
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 8.283128515447979e-06,
+      "loss": 0.0715,
+      "step": 12981
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 8.28299522832085e-06,
+      "loss": 0.0908,
+      "step": 12982
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 8.28286194119372e-06,
+      "loss": 0.0903,
+      "step": 12983
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 8.282728654066591e-06,
+      "loss": 0.121,
+      "step": 12984
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 8.282595366939462e-06,
+      "loss": 0.1139,
+      "step": 12985
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 8.282462079812332e-06,
+      "loss": 0.0969,
+      "step": 12986
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 8.282328792685203e-06,
+      "loss": 0.1036,
+      "step": 12987
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 8.282195505558074e-06,
+      "loss": 0.1171,
+      "step": 12988
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 8.282062218430945e-06,
+      "loss": 0.1095,
+      "step": 12989
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 8.281928931303816e-06,
+      "loss": 0.1122,
+      "step": 12990
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 8.281795644176685e-06,
+      "loss": 0.111,
+      "step": 12991
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 8.281662357049556e-06,
+      "loss": 0.1082,
+      "step": 12992
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 8.281529069922427e-06,
+      "loss": 0.0814,
+      "step": 12993
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 8.281395782795298e-06,
+      "loss": 0.1342,
+      "step": 12994
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 8.281262495668169e-06,
+      "loss": 0.0948,
+      "step": 12995
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 8.28112920854104e-06,
+      "loss": 0.1069,
+      "step": 12996
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 8.28099592141391e-06,
+      "loss": 0.1251,
+      "step": 12997
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 8.280862634286782e-06,
+      "loss": 0.0727,
+      "step": 12998
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 8.280729347159651e-06,
+      "loss": 0.1264,
+      "step": 12999
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 8.280596060032522e-06,
+      "loss": 0.0976,
+      "step": 13000
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 8.280462772905393e-06,
+      "loss": 0.1202,
+      "step": 13001
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 8.280329485778264e-06,
+      "loss": 0.1091,
+      "step": 13002
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 8.280196198651135e-06,
+      "loss": 0.1058,
+      "step": 13003
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 8.280062911524006e-06,
+      "loss": 0.1527,
+      "step": 13004
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 8.279929624396876e-06,
+      "loss": 0.082,
+      "step": 13005
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 8.279796337269748e-06,
+      "loss": 0.0585,
+      "step": 13006
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 8.279663050142618e-06,
+      "loss": 0.1348,
+      "step": 13007
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 8.279529763015489e-06,
+      "loss": 0.0978,
+      "step": 13008
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 8.27939647588836e-06,
+      "loss": 0.1721,
+      "step": 13009
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 8.279263188761229e-06,
+      "loss": 0.0977,
+      "step": 13010
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 8.279129901634102e-06,
+      "loss": 0.1452,
+      "step": 13011
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 8.278996614506971e-06,
+      "loss": 0.114,
+      "step": 13012
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 8.278863327379842e-06,
+      "loss": 0.1029,
+      "step": 13013
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 8.278730040252713e-06,
+      "loss": 0.0832,
+      "step": 13014
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 8.278596753125584e-06,
+      "loss": 0.14,
+      "step": 13015
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 8.278463465998455e-06,
+      "loss": 0.0857,
+      "step": 13016
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 8.278330178871326e-06,
+      "loss": 0.1124,
+      "step": 13017
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 8.278196891744195e-06,
+      "loss": 0.0863,
+      "step": 13018
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 8.278063604617068e-06,
+      "loss": 0.1018,
+      "step": 13019
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 8.277930317489937e-06,
+      "loss": 0.099,
+      "step": 13020
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 8.277797030362808e-06,
+      "loss": 0.1117,
+      "step": 13021
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 8.277663743235679e-06,
+      "loss": 0.0928,
+      "step": 13022
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 8.27753045610855e-06,
+      "loss": 0.0932,
+      "step": 13023
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 8.277397168981421e-06,
+      "loss": 0.1414,
+      "step": 13024
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 8.277263881854292e-06,
+      "loss": 0.0834,
+      "step": 13025
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 8.277130594727161e-06,
+      "loss": 0.1149,
+      "step": 13026
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 8.276997307600032e-06,
+      "loss": 0.0505,
+      "step": 13027
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 8.276864020472903e-06,
+      "loss": 0.0905,
+      "step": 13028
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 8.276730733345774e-06,
+      "loss": 0.1164,
+      "step": 13029
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 8.276597446218645e-06,
+      "loss": 0.0679,
+      "step": 13030
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 8.276464159091515e-06,
+      "loss": 0.1133,
+      "step": 13031
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 8.276330871964386e-06,
+      "loss": 0.0961,
+      "step": 13032
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 8.276197584837257e-06,
+      "loss": 0.1164,
+      "step": 13033
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 8.276064297710128e-06,
+      "loss": 0.0702,
+      "step": 13034
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 8.275931010582999e-06,
+      "loss": 0.1241,
+      "step": 13035
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 8.27579772345587e-06,
+      "loss": 0.0805,
+      "step": 13036
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 8.275664436328739e-06,
+      "loss": 0.0798,
+      "step": 13037
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 8.275531149201612e-06,
+      "loss": 0.092,
+      "step": 13038
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 8.275397862074481e-06,
+      "loss": 0.0926,
+      "step": 13039
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 8.275264574947352e-06,
+      "loss": 0.1174,
+      "step": 13040
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 8.275131287820223e-06,
+      "loss": 0.0952,
+      "step": 13041
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 8.274998000693094e-06,
+      "loss": 0.0848,
+      "step": 13042
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 8.274864713565965e-06,
+      "loss": 0.1086,
+      "step": 13043
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 8.274731426438836e-06,
+      "loss": 0.1048,
+      "step": 13044
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 8.274598139311705e-06,
+      "loss": 0.0878,
+      "step": 13045
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 8.274464852184578e-06,
+      "loss": 0.0906,
+      "step": 13046
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 8.274331565057447e-06,
+      "loss": 0.0988,
+      "step": 13047
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 8.274198277930318e-06,
+      "loss": 0.0891,
+      "step": 13048
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 8.27406499080319e-06,
+      "loss": 0.1135,
+      "step": 13049
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 8.273931703676059e-06,
+      "loss": 0.0971,
+      "step": 13050
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 8.273798416548931e-06,
+      "loss": 0.1377,
+      "step": 13051
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 8.2736651294218e-06,
+      "loss": 0.0819,
+      "step": 13052
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 8.273531842294671e-06,
+      "loss": 0.1182,
+      "step": 13053
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 8.273398555167542e-06,
+      "loss": 0.068,
+      "step": 13054
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 8.273265268040413e-06,
+      "loss": 0.0753,
+      "step": 13055
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 8.273131980913284e-06,
+      "loss": 0.059,
+      "step": 13056
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 8.272998693786155e-06,
+      "loss": 0.1084,
+      "step": 13057
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 8.272865406659025e-06,
+      "loss": 0.1076,
+      "step": 13058
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 8.272732119531896e-06,
+      "loss": 0.1239,
+      "step": 13059
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 8.272598832404767e-06,
+      "loss": 0.0986,
+      "step": 13060
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 8.272465545277638e-06,
+      "loss": 0.1605,
+      "step": 13061
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 8.272332258150509e-06,
+      "loss": 0.1101,
+      "step": 13062
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 8.27219897102338e-06,
+      "loss": 0.0928,
+      "step": 13063
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 8.272065683896249e-06,
+      "loss": 0.1301,
+      "step": 13064
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 8.271932396769122e-06,
+      "loss": 0.12,
+      "step": 13065
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 8.271799109641991e-06,
+      "loss": 0.0957,
+      "step": 13066
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 8.271665822514862e-06,
+      "loss": 0.1214,
+      "step": 13067
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 8.271532535387733e-06,
+      "loss": 0.1306,
+      "step": 13068
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 8.271399248260602e-06,
+      "loss": 0.1005,
+      "step": 13069
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 8.271265961133475e-06,
+      "loss": 0.1128,
+      "step": 13070
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 8.271132674006344e-06,
+      "loss": 0.1175,
+      "step": 13071
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 8.270999386879215e-06,
+      "loss": 0.0924,
+      "step": 13072
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 8.270866099752086e-06,
+      "loss": 0.1001,
+      "step": 13073
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 8.270732812624957e-06,
+      "loss": 0.0871,
+      "step": 13074
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 8.270599525497828e-06,
+      "loss": 0.1202,
+      "step": 13075
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 8.2704662383707e-06,
+      "loss": 0.1289,
+      "step": 13076
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 8.270332951243569e-06,
+      "loss": 0.104,
+      "step": 13077
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 8.270199664116441e-06,
+      "loss": 0.0676,
+      "step": 13078
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 8.27006637698931e-06,
+      "loss": 0.09,
+      "step": 13079
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 8.269933089862182e-06,
+      "loss": 0.143,
+      "step": 13080
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 8.269799802735053e-06,
+      "loss": 0.1179,
+      "step": 13081
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 8.269666515607924e-06,
+      "loss": 0.1111,
+      "step": 13082
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 8.269533228480795e-06,
+      "loss": 0.0864,
+      "step": 13083
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 8.269399941353666e-06,
+      "loss": 0.0914,
+      "step": 13084
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 8.269266654226535e-06,
+      "loss": 0.1106,
+      "step": 13085
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 8.269133367099408e-06,
+      "loss": 0.1168,
+      "step": 13086
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 8.269000079972277e-06,
+      "loss": 0.0833,
+      "step": 13087
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 8.268866792845148e-06,
+      "loss": 0.0667,
+      "step": 13088
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 8.268733505718019e-06,
+      "loss": 0.1377,
+      "step": 13089
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 8.268600218590888e-06,
+      "loss": 0.0954,
+      "step": 13090
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 8.26846693146376e-06,
+      "loss": 0.1097,
+      "step": 13091
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 8.26833364433663e-06,
+      "loss": 0.1452,
+      "step": 13092
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 8.268200357209501e-06,
+      "loss": 0.1163,
+      "step": 13093
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 8.268067070082372e-06,
+      "loss": 0.0957,
+      "step": 13094
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 8.267933782955243e-06,
+      "loss": 0.1189,
+      "step": 13095
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 8.267800495828114e-06,
+      "loss": 0.086,
+      "step": 13096
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 8.267667208700985e-06,
+      "loss": 0.0939,
+      "step": 13097
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 8.267533921573854e-06,
+      "loss": 0.1402,
+      "step": 13098
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 8.267400634446725e-06,
+      "loss": 0.0857,
+      "step": 13099
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 8.267267347319596e-06,
+      "loss": 0.0763,
+      "step": 13100
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 8.267134060192467e-06,
+      "loss": 0.1066,
+      "step": 13101
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 8.267000773065338e-06,
+      "loss": 0.1069,
+      "step": 13102
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 8.26686748593821e-06,
+      "loss": 0.0759,
+      "step": 13103
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 8.266734198811079e-06,
+      "loss": 0.0915,
+      "step": 13104
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 8.266600911683951e-06,
+      "loss": 0.0893,
+      "step": 13105
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 8.26646762455682e-06,
+      "loss": 0.0936,
+      "step": 13106
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 8.266334337429692e-06,
+      "loss": 0.0655,
+      "step": 13107
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 8.266201050302563e-06,
+      "loss": 0.1226,
+      "step": 13108
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 8.266067763175432e-06,
+      "loss": 0.1299,
+      "step": 13109
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 8.265934476048305e-06,
+      "loss": 0.1177,
+      "step": 13110
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 8.265801188921174e-06,
+      "loss": 0.1177,
+      "step": 13111
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 8.265667901794045e-06,
+      "loss": 0.1025,
+      "step": 13112
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 8.265534614666916e-06,
+      "loss": 0.1246,
+      "step": 13113
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 8.265401327539787e-06,
+      "loss": 0.1172,
+      "step": 13114
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 8.265268040412658e-06,
+      "loss": 0.1612,
+      "step": 13115
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 8.265134753285529e-06,
+      "loss": 0.1173,
+      "step": 13116
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 8.265001466158398e-06,
+      "loss": 0.1129,
+      "step": 13117
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 8.264868179031271e-06,
+      "loss": 0.1212,
+      "step": 13118
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 8.26473489190414e-06,
+      "loss": 0.1126,
+      "step": 13119
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 8.264601604777011e-06,
+      "loss": 0.1085,
+      "step": 13120
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 8.264468317649882e-06,
+      "loss": 0.1062,
+      "step": 13121
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 8.264335030522753e-06,
+      "loss": 0.1173,
+      "step": 13122
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 8.264201743395624e-06,
+      "loss": 0.1431,
+      "step": 13123
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 8.264068456268495e-06,
+      "loss": 0.1067,
+      "step": 13124
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 8.263935169141364e-06,
+      "loss": 0.097,
+      "step": 13125
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 8.263801882014235e-06,
+      "loss": 0.1204,
+      "step": 13126
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 8.263668594887106e-06,
+      "loss": 0.1089,
+      "step": 13127
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 8.263535307759977e-06,
+      "loss": 0.0984,
+      "step": 13128
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 8.263402020632848e-06,
+      "loss": 0.0784,
+      "step": 13129
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 8.263268733505718e-06,
+      "loss": 0.0706,
+      "step": 13130
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 8.263135446378589e-06,
+      "loss": 0.0992,
+      "step": 13131
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 8.26300215925146e-06,
+      "loss": 0.0924,
+      "step": 13132
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 8.26286887212433e-06,
+      "loss": 0.1326,
+      "step": 13133
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 8.262735584997202e-06,
+      "loss": 0.1418,
+      "step": 13134
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 8.262602297870073e-06,
+      "loss": 0.1075,
+      "step": 13135
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 8.262469010742942e-06,
+      "loss": 0.0647,
+      "step": 13136
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 8.262335723615815e-06,
+      "loss": 0.1444,
+      "step": 13137
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 8.262202436488684e-06,
+      "loss": 0.1087,
+      "step": 13138
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 8.262069149361555e-06,
+      "loss": 0.0914,
+      "step": 13139
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 8.261935862234426e-06,
+      "loss": 0.1314,
+      "step": 13140
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 8.261802575107297e-06,
+      "loss": 0.0674,
+      "step": 13141
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 8.261669287980168e-06,
+      "loss": 0.1142,
+      "step": 13142
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 8.261536000853039e-06,
+      "loss": 0.1219,
+      "step": 13143
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 8.261402713725908e-06,
+      "loss": 0.0978,
+      "step": 13144
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 8.261269426598781e-06,
+      "loss": 0.0712,
+      "step": 13145
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 8.26113613947165e-06,
+      "loss": 0.1062,
+      "step": 13146
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 8.261002852344521e-06,
+      "loss": 0.1522,
+      "step": 13147
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 8.260869565217392e-06,
+      "loss": 0.0804,
+      "step": 13148
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 8.260736278090262e-06,
+      "loss": 0.0642,
+      "step": 13149
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 8.260602990963134e-06,
+      "loss": 0.0976,
+      "step": 13150
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 8.260469703836004e-06,
+      "loss": 0.1208,
+      "step": 13151
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 8.260336416708875e-06,
+      "loss": 0.0861,
+      "step": 13152
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 8.260203129581746e-06,
+      "loss": 0.1042,
+      "step": 13153
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 8.260069842454617e-06,
+      "loss": 0.1576,
+      "step": 13154
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 8.259936555327488e-06,
+      "loss": 0.0934,
+      "step": 13155
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 8.259803268200359e-06,
+      "loss": 0.1222,
+      "step": 13156
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 8.259669981073228e-06,
+      "loss": 0.1082,
+      "step": 13157
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 8.2595366939461e-06,
+      "loss": 0.1323,
+      "step": 13158
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 8.25940340681897e-06,
+      "loss": 0.0889,
+      "step": 13159
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 8.25927011969184e-06,
+      "loss": 0.1166,
+      "step": 13160
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 8.259136832564712e-06,
+      "loss": 0.0645,
+      "step": 13161
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 8.259003545437583e-06,
+      "loss": 0.0672,
+      "step": 13162
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 8.258870258310454e-06,
+      "loss": 0.0922,
+      "step": 13163
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 8.258736971183325e-06,
+      "loss": 0.1077,
+      "step": 13164
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 8.258603684056194e-06,
+      "loss": 0.0754,
+      "step": 13165
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 8.258470396929065e-06,
+      "loss": 0.1256,
+      "step": 13166
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 8.258337109801936e-06,
+      "loss": 0.1103,
+      "step": 13167
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 8.258203822674807e-06,
+      "loss": 0.0681,
+      "step": 13168
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 8.258070535547678e-06,
+      "loss": 0.0824,
+      "step": 13169
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 8.257937248420547e-06,
+      "loss": 0.1127,
+      "step": 13170
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 8.257803961293418e-06,
+      "loss": 0.1137,
+      "step": 13171
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 8.25767067416629e-06,
+      "loss": 0.1095,
+      "step": 13172
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 8.25753738703916e-06,
+      "loss": 0.1246,
+      "step": 13173
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 8.257404099912031e-06,
+      "loss": 0.1044,
+      "step": 13174
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 8.257270812784902e-06,
+      "loss": 0.0752,
+      "step": 13175
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 8.257137525657772e-06,
+      "loss": 0.1363,
+      "step": 13176
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 8.257004238530644e-06,
+      "loss": 0.0934,
+      "step": 13177
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 8.256870951403514e-06,
+      "loss": 0.0995,
+      "step": 13178
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 8.256737664276385e-06,
+      "loss": 0.1429,
+      "step": 13179
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 8.256604377149256e-06,
+      "loss": 0.1148,
+      "step": 13180
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 8.256471090022127e-06,
+      "loss": 0.1284,
+      "step": 13181
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 8.256337802894998e-06,
+      "loss": 0.0741,
+      "step": 13182
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 8.256204515767869e-06,
+      "loss": 0.0789,
+      "step": 13183
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 8.256071228640738e-06,
+      "loss": 0.1029,
+      "step": 13184
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 8.25593794151361e-06,
+      "loss": 0.1641,
+      "step": 13185
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 8.25580465438648e-06,
+      "loss": 0.0691,
+      "step": 13186
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 8.255671367259351e-06,
+      "loss": 0.0888,
+      "step": 13187
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 8.255538080132222e-06,
+      "loss": 0.156,
+      "step": 13188
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 8.255404793005091e-06,
+      "loss": 0.1026,
+      "step": 13189
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 8.255271505877964e-06,
+      "loss": 0.0799,
+      "step": 13190
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 8.255138218750833e-06,
+      "loss": 0.0677,
+      "step": 13191
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 8.255004931623704e-06,
+      "loss": 0.1052,
+      "step": 13192
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 8.254871644496575e-06,
+      "loss": 0.1343,
+      "step": 13193
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 8.254738357369446e-06,
+      "loss": 0.0815,
+      "step": 13194
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 8.254605070242317e-06,
+      "loss": 0.1118,
+      "step": 13195
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 8.254471783115188e-06,
+      "loss": 0.1094,
+      "step": 13196
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 8.254338495988057e-06,
+      "loss": 0.105,
+      "step": 13197
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 8.254205208860928e-06,
+      "loss": 0.0911,
+      "step": 13198
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 8.2540719217338e-06,
+      "loss": 0.1068,
+      "step": 13199
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 8.25393863460667e-06,
+      "loss": 0.1373,
+      "step": 13200
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 8.253805347479541e-06,
+      "loss": 0.1093,
+      "step": 13201
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 8.253672060352412e-06,
+      "loss": 0.1762,
+      "step": 13202
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 8.253538773225282e-06,
+      "loss": 0.0808,
+      "step": 13203
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 8.253405486098154e-06,
+      "loss": 0.108,
+      "step": 13204
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 8.253272198971024e-06,
+      "loss": 0.0993,
+      "step": 13205
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 8.253138911843895e-06,
+      "loss": 0.1567,
+      "step": 13206
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 8.253005624716766e-06,
+      "loss": 0.0874,
+      "step": 13207
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 8.252872337589635e-06,
+      "loss": 0.1068,
+      "step": 13208
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 8.252739050462508e-06,
+      "loss": 0.1167,
+      "step": 13209
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 8.252605763335377e-06,
+      "loss": 0.0724,
+      "step": 13210
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 8.252472476208248e-06,
+      "loss": 0.1035,
+      "step": 13211
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 8.252339189081119e-06,
+      "loss": 0.1055,
+      "step": 13212
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 8.25220590195399e-06,
+      "loss": 0.1393,
+      "step": 13213
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 8.252072614826861e-06,
+      "loss": 0.0906,
+      "step": 13214
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 8.251939327699732e-06,
+      "loss": 0.0843,
+      "step": 13215
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 8.251806040572601e-06,
+      "loss": 0.1138,
+      "step": 13216
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 8.251672753445474e-06,
+      "loss": 0.0868,
+      "step": 13217
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 8.251539466318343e-06,
+      "loss": 0.1068,
+      "step": 13218
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 8.251406179191214e-06,
+      "loss": 0.1094,
+      "step": 13219
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 8.251272892064085e-06,
+      "loss": 0.0757,
+      "step": 13220
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 8.251139604936956e-06,
+      "loss": 0.1159,
+      "step": 13221
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 8.251006317809827e-06,
+      "loss": 0.0974,
+      "step": 13222
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 8.250873030682698e-06,
+      "loss": 0.0689,
+      "step": 13223
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 8.250739743555568e-06,
+      "loss": 0.1031,
+      "step": 13224
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 8.250606456428439e-06,
+      "loss": 0.1128,
+      "step": 13225
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 8.25047316930131e-06,
+      "loss": 0.1222,
+      "step": 13226
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 8.25033988217418e-06,
+      "loss": 0.1139,
+      "step": 13227
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 8.250206595047052e-06,
+      "loss": 0.0957,
+      "step": 13228
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 8.25007330791992e-06,
+      "loss": 0.1173,
+      "step": 13229
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 8.249940020792792e-06,
+      "loss": 0.1005,
+      "step": 13230
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 8.249806733665663e-06,
+      "loss": 0.1053,
+      "step": 13231
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 8.249673446538534e-06,
+      "loss": 0.1068,
+      "step": 13232
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 8.249540159411405e-06,
+      "loss": 0.0943,
+      "step": 13233
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 8.249406872284276e-06,
+      "loss": 0.0842,
+      "step": 13234
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 8.249273585157145e-06,
+      "loss": 0.1149,
+      "step": 13235
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 8.249140298030018e-06,
+      "loss": 0.0773,
+      "step": 13236
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 8.249007010902887e-06,
+      "loss": 0.1311,
+      "step": 13237
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 8.248873723775758e-06,
+      "loss": 0.1284,
+      "step": 13238
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 8.248740436648629e-06,
+      "loss": 0.1174,
+      "step": 13239
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 8.2486071495215e-06,
+      "loss": 0.1439,
+      "step": 13240
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 8.248473862394371e-06,
+      "loss": 0.0766,
+      "step": 13241
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 8.248340575267242e-06,
+      "loss": 0.0854,
+      "step": 13242
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 8.248207288140111e-06,
+      "loss": 0.0903,
+      "step": 13243
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 8.248074001012984e-06,
+      "loss": 0.0574,
+      "step": 13244
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 8.247940713885853e-06,
+      "loss": 0.092,
+      "step": 13245
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 8.247807426758724e-06,
+      "loss": 0.0706,
+      "step": 13246
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 8.247674139631595e-06,
+      "loss": 0.0868,
+      "step": 13247
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 8.247540852504465e-06,
+      "loss": 0.1353,
+      "step": 13248
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 8.247407565377337e-06,
+      "loss": 0.1062,
+      "step": 13249
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 8.247274278250207e-06,
+      "loss": 0.1126,
+      "step": 13250
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 8.247140991123078e-06,
+      "loss": 0.0668,
+      "step": 13251
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 8.247007703995949e-06,
+      "loss": 0.1011,
+      "step": 13252
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 8.24687441686882e-06,
+      "loss": 0.0901,
+      "step": 13253
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 8.24674112974169e-06,
+      "loss": 0.0789,
+      "step": 13254
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 8.246607842614562e-06,
+      "loss": 0.0982,
+      "step": 13255
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 8.246474555487431e-06,
+      "loss": 0.096,
+      "step": 13256
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 8.246341268360304e-06,
+      "loss": 0.0848,
+      "step": 13257
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 8.246207981233173e-06,
+      "loss": 0.0951,
+      "step": 13258
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 8.246074694106044e-06,
+      "loss": 0.0783,
+      "step": 13259
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 8.245941406978915e-06,
+      "loss": 0.0894,
+      "step": 13260
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 8.245808119851786e-06,
+      "loss": 0.1129,
+      "step": 13261
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 8.245674832724657e-06,
+      "loss": 0.1429,
+      "step": 13262
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 8.245541545597528e-06,
+      "loss": 0.1143,
+      "step": 13263
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 8.245408258470397e-06,
+      "loss": 0.0815,
+      "step": 13264
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 8.245274971343268e-06,
+      "loss": 0.1118,
+      "step": 13265
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 8.24514168421614e-06,
+      "loss": 0.0815,
+      "step": 13266
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 8.24500839708901e-06,
+      "loss": 0.0896,
+      "step": 13267
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 8.244875109961881e-06,
+      "loss": 0.1003,
+      "step": 13268
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 8.24474182283475e-06,
+      "loss": 0.1247,
+      "step": 13269
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 8.244608535707621e-06,
+      "loss": 0.08,
+      "step": 13270
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 8.244475248580492e-06,
+      "loss": 0.1012,
+      "step": 13271
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 8.244341961453363e-06,
+      "loss": 0.0878,
+      "step": 13272
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 8.244208674326234e-06,
+      "loss": 0.1184,
+      "step": 13273
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 8.244075387199105e-06,
+      "loss": 0.1058,
+      "step": 13274
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 8.243942100071975e-06,
+      "loss": 0.1077,
+      "step": 13275
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 8.243808812944847e-06,
+      "loss": 0.0958,
+      "step": 13276
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 8.243675525817717e-06,
+      "loss": 0.0811,
+      "step": 13277
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 8.243542238690588e-06,
+      "loss": 0.0989,
+      "step": 13278
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 8.243408951563459e-06,
+      "loss": 0.1445,
+      "step": 13279
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 8.24327566443633e-06,
+      "loss": 0.0822,
+      "step": 13280
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 8.2431423773092e-06,
+      "loss": 0.1077,
+      "step": 13281
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 8.243009090182072e-06,
+      "loss": 0.1295,
+      "step": 13282
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 8.242875803054941e-06,
+      "loss": 0.1168,
+      "step": 13283
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 8.242742515927812e-06,
+      "loss": 0.0694,
+      "step": 13284
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 8.242609228800683e-06,
+      "loss": 0.1063,
+      "step": 13285
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 8.242475941673554e-06,
+      "loss": 0.0876,
+      "step": 13286
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 8.242342654546425e-06,
+      "loss": 0.1158,
+      "step": 13287
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 8.242209367419294e-06,
+      "loss": 0.1247,
+      "step": 13288
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 8.242076080292167e-06,
+      "loss": 0.1363,
+      "step": 13289
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 8.241942793165036e-06,
+      "loss": 0.1337,
+      "step": 13290
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 8.241809506037907e-06,
+      "loss": 0.0916,
+      "step": 13291
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 8.241676218910778e-06,
+      "loss": 0.138,
+      "step": 13292
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 8.24154293178365e-06,
+      "loss": 0.0798,
+      "step": 13293
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 8.24140964465652e-06,
+      "loss": 0.0894,
+      "step": 13294
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 8.241276357529391e-06,
+      "loss": 0.1481,
+      "step": 13295
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 8.24114307040226e-06,
+      "loss": 0.0983,
+      "step": 13296
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 8.241009783275132e-06,
+      "loss": 0.0739,
+      "step": 13297
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 8.240876496148003e-06,
+      "loss": 0.0828,
+      "step": 13298
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 8.240743209020874e-06,
+      "loss": 0.099,
+      "step": 13299
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 8.240609921893745e-06,
+      "loss": 0.0799,
+      "step": 13300
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 8.240476634766616e-06,
+      "loss": 0.109,
+      "step": 13301
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 8.240343347639485e-06,
+      "loss": 0.0814,
+      "step": 13302
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 8.240210060512357e-06,
+      "loss": 0.1104,
+      "step": 13303
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 8.240076773385227e-06,
+      "loss": 0.0616,
+      "step": 13304
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 8.239943486258098e-06,
+      "loss": 0.1285,
+      "step": 13305
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 8.239810199130969e-06,
+      "loss": 0.1031,
+      "step": 13306
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 8.239676912003838e-06,
+      "loss": 0.0739,
+      "step": 13307
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 8.23954362487671e-06,
+      "loss": 0.084,
+      "step": 13308
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 8.23941033774958e-06,
+      "loss": 0.0987,
+      "step": 13309
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 8.239277050622451e-06,
+      "loss": 0.0975,
+      "step": 13310
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 8.239143763495322e-06,
+      "loss": 0.1165,
+      "step": 13311
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 8.239010476368193e-06,
+      "loss": 0.1547,
+      "step": 13312
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 8.238877189241064e-06,
+      "loss": 0.1076,
+      "step": 13313
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 8.238743902113935e-06,
+      "loss": 0.1153,
+      "step": 13314
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 8.238610614986804e-06,
+      "loss": 0.0893,
+      "step": 13315
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 8.238477327859677e-06,
+      "loss": 0.0649,
+      "step": 13316
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 8.238344040732546e-06,
+      "loss": 0.1269,
+      "step": 13317
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 8.238210753605417e-06,
+      "loss": 0.097,
+      "step": 13318
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 8.238077466478288e-06,
+      "loss": 0.1319,
+      "step": 13319
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 8.23794417935116e-06,
+      "loss": 0.1059,
+      "step": 13320
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 8.23781089222403e-06,
+      "loss": 0.0828,
+      "step": 13321
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 8.237677605096901e-06,
+      "loss": 0.0689,
+      "step": 13322
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 8.23754431796977e-06,
+      "loss": 0.0751,
+      "step": 13323
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 8.237411030842642e-06,
+      "loss": 0.0992,
+      "step": 13324
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 8.237277743715513e-06,
+      "loss": 0.099,
+      "step": 13325
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 8.237144456588384e-06,
+      "loss": 0.1083,
+      "step": 13326
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 8.237011169461255e-06,
+      "loss": 0.0939,
+      "step": 13327
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 8.236877882334124e-06,
+      "loss": 0.0905,
+      "step": 13328
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 8.236744595206997e-06,
+      "loss": 0.0799,
+      "step": 13329
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 8.236611308079866e-06,
+      "loss": 0.112,
+      "step": 13330
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 8.236478020952737e-06,
+      "loss": 0.0825,
+      "step": 13331
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 8.236344733825608e-06,
+      "loss": 0.1161,
+      "step": 13332
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 8.236211446698479e-06,
+      "loss": 0.1201,
+      "step": 13333
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 8.23607815957135e-06,
+      "loss": 0.0996,
+      "step": 13334
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 8.235944872444221e-06,
+      "loss": 0.1008,
+      "step": 13335
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 8.23581158531709e-06,
+      "loss": 0.1075,
+      "step": 13336
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 8.235678298189961e-06,
+      "loss": 0.0845,
+      "step": 13337
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 8.235545011062832e-06,
+      "loss": 0.1168,
+      "step": 13338
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 8.235411723935703e-06,
+      "loss": 0.0807,
+      "step": 13339
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 8.235278436808574e-06,
+      "loss": 0.0961,
+      "step": 13340
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 8.235145149681445e-06,
+      "loss": 0.1179,
+      "step": 13341
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 8.235011862554314e-06,
+      "loss": 0.0999,
+      "step": 13342
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 8.234878575427185e-06,
+      "loss": 0.0951,
+      "step": 13343
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 8.234745288300056e-06,
+      "loss": 0.1166,
+      "step": 13344
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 8.234612001172927e-06,
+      "loss": 0.1167,
+      "step": 13345
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 8.234478714045798e-06,
+      "loss": 0.1346,
+      "step": 13346
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 8.234345426918668e-06,
+      "loss": 0.0863,
+      "step": 13347
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 8.23421213979154e-06,
+      "loss": 0.1257,
+      "step": 13348
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 8.23407885266441e-06,
+      "loss": 0.0838,
+      "step": 13349
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 8.23394556553728e-06,
+      "loss": 0.1394,
+      "step": 13350
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 8.233812278410152e-06,
+      "loss": 0.1116,
+      "step": 13351
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 8.233678991283023e-06,
+      "loss": 0.0751,
+      "step": 13352
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 8.233545704155894e-06,
+      "loss": 0.0579,
+      "step": 13353
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 8.233412417028765e-06,
+      "loss": 0.0872,
+      "step": 13354
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 8.233279129901634e-06,
+      "loss": 0.0845,
+      "step": 13355
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 8.233145842774507e-06,
+      "loss": 0.0986,
+      "step": 13356
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 8.233012555647376e-06,
+      "loss": 0.0787,
+      "step": 13357
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 8.232879268520247e-06,
+      "loss": 0.1144,
+      "step": 13358
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 8.232745981393118e-06,
+      "loss": 0.1348,
+      "step": 13359
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 8.232612694265989e-06,
+      "loss": 0.1058,
+      "step": 13360
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 8.23247940713886e-06,
+      "loss": 0.0814,
+      "step": 13361
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 8.23234612001173e-06,
+      "loss": 0.0864,
+      "step": 13362
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 8.2322128328846e-06,
+      "loss": 0.1433,
+      "step": 13363
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 8.232079545757471e-06,
+      "loss": 0.0918,
+      "step": 13364
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 8.231946258630342e-06,
+      "loss": 0.1236,
+      "step": 13365
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 8.231812971503213e-06,
+      "loss": 0.1033,
+      "step": 13366
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 8.231679684376084e-06,
+      "loss": 0.0921,
+      "step": 13367
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 8.231546397248954e-06,
+      "loss": 0.1093,
+      "step": 13368
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 8.231413110121825e-06,
+      "loss": 0.0763,
+      "step": 13369
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 8.231279822994696e-06,
+      "loss": 0.0919,
+      "step": 13370
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 8.231146535867566e-06,
+      "loss": 0.1251,
+      "step": 13371
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 8.231013248740437e-06,
+      "loss": 0.1463,
+      "step": 13372
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 8.230879961613308e-06,
+      "loss": 0.1169,
+      "step": 13373
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 8.230746674486178e-06,
+      "loss": 0.1018,
+      "step": 13374
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 8.23061338735905e-06,
+      "loss": 0.0694,
+      "step": 13375
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 8.23048010023192e-06,
+      "loss": 0.0789,
+      "step": 13376
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 8.23034681310479e-06,
+      "loss": 0.0761,
+      "step": 13377
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 8.230213525977662e-06,
+      "loss": 0.1461,
+      "step": 13378
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 8.230080238850533e-06,
+      "loss": 0.1253,
+      "step": 13379
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 8.229946951723404e-06,
+      "loss": 0.0882,
+      "step": 13380
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 8.229813664596275e-06,
+      "loss": 0.0865,
+      "step": 13381
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 8.229680377469144e-06,
+      "loss": 0.1008,
+      "step": 13382
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 8.229547090342015e-06,
+      "loss": 0.0929,
+      "step": 13383
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 8.229413803214886e-06,
+      "loss": 0.1041,
+      "step": 13384
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 8.229280516087757e-06,
+      "loss": 0.1056,
+      "step": 13385
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 8.229147228960628e-06,
+      "loss": 0.1053,
+      "step": 13386
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 8.229013941833497e-06,
+      "loss": 0.0951,
+      "step": 13387
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 8.22888065470637e-06,
+      "loss": 0.0762,
+      "step": 13388
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 8.22874736757924e-06,
+      "loss": 0.0916,
+      "step": 13389
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 8.22861408045211e-06,
+      "loss": 0.0964,
+      "step": 13390
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 8.228480793324981e-06,
+      "loss": 0.1133,
+      "step": 13391
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 8.228347506197852e-06,
+      "loss": 0.1203,
+      "step": 13392
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 8.228214219070723e-06,
+      "loss": 0.0763,
+      "step": 13393
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 8.228080931943594e-06,
+      "loss": 0.1129,
+      "step": 13394
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 8.227947644816464e-06,
+      "loss": 0.136,
+      "step": 13395
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 8.227814357689336e-06,
+      "loss": 0.075,
+      "step": 13396
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 8.227681070562206e-06,
+      "loss": 0.0725,
+      "step": 13397
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 8.227547783435077e-06,
+      "loss": 0.1068,
+      "step": 13398
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 8.227414496307948e-06,
+      "loss": 0.0842,
+      "step": 13399
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 8.227281209180819e-06,
+      "loss": 0.11,
+      "step": 13400
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 8.22714792205369e-06,
+      "loss": 0.1178,
+      "step": 13401
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 8.227014634926559e-06,
+      "loss": 0.0992,
+      "step": 13402
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 8.22688134779943e-06,
+      "loss": 0.1151,
+      "step": 13403
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 8.226748060672301e-06,
+      "loss": 0.0717,
+      "step": 13404
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 8.226614773545172e-06,
+      "loss": 0.0582,
+      "step": 13405
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 8.226481486418043e-06,
+      "loss": 0.1284,
+      "step": 13406
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 8.226348199290914e-06,
+      "loss": 0.1269,
+      "step": 13407
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 8.226214912163783e-06,
+      "loss": 0.1015,
+      "step": 13408
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 8.226081625036654e-06,
+      "loss": 0.1335,
+      "step": 13409
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 8.225948337909525e-06,
+      "loss": 0.1201,
+      "step": 13410
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 8.225815050782396e-06,
+      "loss": 0.0762,
+      "step": 13411
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 8.225681763655267e-06,
+      "loss": 0.0847,
+      "step": 13412
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 8.225548476528138e-06,
+      "loss": 0.1033,
+      "step": 13413
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 8.225415189401007e-06,
+      "loss": 0.1198,
+      "step": 13414
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 8.22528190227388e-06,
+      "loss": 0.0829,
+      "step": 13415
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 8.22514861514675e-06,
+      "loss": 0.092,
+      "step": 13416
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 8.22501532801962e-06,
+      "loss": 0.0745,
+      "step": 13417
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 8.224882040892491e-06,
+      "loss": 0.1208,
+      "step": 13418
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 8.224748753765362e-06,
+      "loss": 0.0971,
+      "step": 13419
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 8.224615466638233e-06,
+      "loss": 0.0723,
+      "step": 13420
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 8.224482179511103e-06,
+      "loss": 0.103,
+      "step": 13421
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 8.224348892383974e-06,
+      "loss": 0.1204,
+      "step": 13422
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 8.224215605256845e-06,
+      "loss": 0.1027,
+      "step": 13423
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 8.224082318129716e-06,
+      "loss": 0.0644,
+      "step": 13424
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 8.223949031002587e-06,
+      "loss": 0.1496,
+      "step": 13425
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 8.223815743875458e-06,
+      "loss": 0.1197,
+      "step": 13426
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 8.223682456748327e-06,
+      "loss": 0.1134,
+      "step": 13427
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 8.2235491696212e-06,
+      "loss": 0.0905,
+      "step": 13428
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 8.223415882494069e-06,
+      "loss": 0.0835,
+      "step": 13429
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 8.22328259536694e-06,
+      "loss": 0.1155,
+      "step": 13430
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 8.223149308239811e-06,
+      "loss": 0.1239,
+      "step": 13431
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 8.223016021112682e-06,
+      "loss": 0.152,
+      "step": 13432
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 8.222882733985553e-06,
+      "loss": 0.0988,
+      "step": 13433
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 8.222749446858424e-06,
+      "loss": 0.0891,
+      "step": 13434
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 8.222616159731293e-06,
+      "loss": 0.1218,
+      "step": 13435
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 8.222482872604164e-06,
+      "loss": 0.162,
+      "step": 13436
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 8.222349585477035e-06,
+      "loss": 0.088,
+      "step": 13437
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 8.222216298349906e-06,
+      "loss": 0.1014,
+      "step": 13438
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 8.222083011222777e-06,
+      "loss": 0.136,
+      "step": 13439
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 8.221949724095648e-06,
+      "loss": 0.1075,
+      "step": 13440
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 8.221816436968517e-06,
+      "loss": 0.1495,
+      "step": 13441
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 8.221683149841388e-06,
+      "loss": 0.1058,
+      "step": 13442
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 8.22154986271426e-06,
+      "loss": 0.0938,
+      "step": 13443
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 8.22141657558713e-06,
+      "loss": 0.0877,
+      "step": 13444
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 8.221283288460001e-06,
+      "loss": 0.1041,
+      "step": 13445
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 8.22115000133287e-06,
+      "loss": 0.108,
+      "step": 13446
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 8.221016714205743e-06,
+      "loss": 0.1002,
+      "step": 13447
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 8.220883427078613e-06,
+      "loss": 0.0976,
+      "step": 13448
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 8.220750139951484e-06,
+      "loss": 0.0692,
+      "step": 13449
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 8.220616852824355e-06,
+      "loss": 0.1335,
+      "step": 13450
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 8.220483565697226e-06,
+      "loss": 0.1068,
+      "step": 13451
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 8.220350278570097e-06,
+      "loss": 0.1205,
+      "step": 13452
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 8.220216991442968e-06,
+      "loss": 0.1121,
+      "step": 13453
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 8.220083704315837e-06,
+      "loss": 0.098,
+      "step": 13454
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 8.21995041718871e-06,
+      "loss": 0.1042,
+      "step": 13455
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 8.219817130061579e-06,
+      "loss": 0.0993,
+      "step": 13456
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 8.21968384293445e-06,
+      "loss": 0.1094,
+      "step": 13457
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 8.219550555807321e-06,
+      "loss": 0.0857,
+      "step": 13458
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 8.219417268680192e-06,
+      "loss": 0.0993,
+      "step": 13459
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 8.219283981553063e-06,
+      "loss": 0.0906,
+      "step": 13460
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 8.219150694425932e-06,
+      "loss": 0.0987,
+      "step": 13461
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 8.219017407298803e-06,
+      "loss": 0.0791,
+      "step": 13462
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 8.218884120171674e-06,
+      "loss": 0.0638,
+      "step": 13463
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 8.218750833044545e-06,
+      "loss": 0.1134,
+      "step": 13464
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 8.218617545917416e-06,
+      "loss": 0.0764,
+      "step": 13465
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 8.218484258790287e-06,
+      "loss": 0.1019,
+      "step": 13466
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 8.218350971663157e-06,
+      "loss": 0.1111,
+      "step": 13467
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 8.218217684536028e-06,
+      "loss": 0.1772,
+      "step": 13468
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 8.218084397408899e-06,
+      "loss": 0.0868,
+      "step": 13469
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 8.21795111028177e-06,
+      "loss": 0.0887,
+      "step": 13470
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 8.21781782315464e-06,
+      "loss": 0.1033,
+      "step": 13471
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 8.217684536027512e-06,
+      "loss": 0.0658,
+      "step": 13472
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 8.217551248900381e-06,
+      "loss": 0.093,
+      "step": 13473
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 8.217417961773254e-06,
+      "loss": 0.1001,
+      "step": 13474
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 8.217284674646123e-06,
+      "loss": 0.0782,
+      "step": 13475
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 8.217151387518994e-06,
+      "loss": 0.1086,
+      "step": 13476
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 8.217018100391865e-06,
+      "loss": 0.1235,
+      "step": 13477
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 8.216884813264736e-06,
+      "loss": 0.1101,
+      "step": 13478
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 8.216751526137607e-06,
+      "loss": 0.1133,
+      "step": 13479
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 8.216618239010476e-06,
+      "loss": 0.0985,
+      "step": 13480
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 8.216484951883347e-06,
+      "loss": 0.1405,
+      "step": 13481
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 8.216351664756218e-06,
+      "loss": 0.1266,
+      "step": 13482
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 8.216218377629089e-06,
+      "loss": 0.1283,
+      "step": 13483
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 8.21608509050196e-06,
+      "loss": 0.1241,
+      "step": 13484
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 8.215951803374831e-06,
+      "loss": 0.1159,
+      "step": 13485
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 8.2158185162477e-06,
+      "loss": 0.1228,
+      "step": 13486
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 8.215685229120573e-06,
+      "loss": 0.0999,
+      "step": 13487
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 8.215551941993442e-06,
+      "loss": 0.1028,
+      "step": 13488
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 8.215418654866313e-06,
+      "loss": 0.0916,
+      "step": 13489
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 8.215285367739184e-06,
+      "loss": 0.0751,
+      "step": 13490
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 8.215152080612055e-06,
+      "loss": 0.1024,
+      "step": 13491
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 8.215018793484926e-06,
+      "loss": 0.0952,
+      "step": 13492
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 8.214885506357797e-06,
+      "loss": 0.107,
+      "step": 13493
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 8.214752219230667e-06,
+      "loss": 0.135,
+      "step": 13494
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 8.21461893210354e-06,
+      "loss": 0.0945,
+      "step": 13495
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 8.214485644976409e-06,
+      "loss": 0.1125,
+      "step": 13496
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 8.21435235784928e-06,
+      "loss": 0.0993,
+      "step": 13497
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 8.21421907072215e-06,
+      "loss": 0.1055,
+      "step": 13498
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 8.21408578359502e-06,
+      "loss": 0.0791,
+      "step": 13499
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 8.213952496467893e-06,
+      "loss": 0.1248,
+      "step": 13500
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 8.213819209340762e-06,
+      "loss": 0.111,
+      "step": 13501
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 8.213685922213633e-06,
+      "loss": 0.1231,
+      "step": 13502
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 8.213552635086504e-06,
+      "loss": 0.0924,
+      "step": 13503
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 8.213419347959375e-06,
+      "loss": 0.1123,
+      "step": 13504
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 8.213286060832246e-06,
+      "loss": 0.091,
+      "step": 13505
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 8.213152773705117e-06,
+      "loss": 0.0806,
+      "step": 13506
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 8.213019486577986e-06,
+      "loss": 0.1164,
+      "step": 13507
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 8.212886199450857e-06,
+      "loss": 0.0995,
+      "step": 13508
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 8.212752912323728e-06,
+      "loss": 0.1379,
+      "step": 13509
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 8.2126196251966e-06,
+      "loss": 0.0884,
+      "step": 13510
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 8.21248633806947e-06,
+      "loss": 0.1358,
+      "step": 13511
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 8.212353050942341e-06,
+      "loss": 0.1135,
+      "step": 13512
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 8.21221976381521e-06,
+      "loss": 0.0901,
+      "step": 13513
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 8.212086476688083e-06,
+      "loss": 0.1071,
+      "step": 13514
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 8.211953189560952e-06,
+      "loss": 0.0981,
+      "step": 13515
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 8.211819902433823e-06,
+      "loss": 0.1052,
+      "step": 13516
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 8.211686615306694e-06,
+      "loss": 0.1149,
+      "step": 13517
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 8.211553328179565e-06,
+      "loss": 0.1129,
+      "step": 13518
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 8.211420041052436e-06,
+      "loss": 0.1392,
+      "step": 13519
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 8.211286753925306e-06,
+      "loss": 0.0838,
+      "step": 13520
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 8.211153466798177e-06,
+      "loss": 0.1382,
+      "step": 13521
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 8.211020179671048e-06,
+      "loss": 0.0754,
+      "step": 13522
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 8.210886892543919e-06,
+      "loss": 0.1128,
+      "step": 13523
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 8.21075360541679e-06,
+      "loss": 0.0837,
+      "step": 13524
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 8.21062031828966e-06,
+      "loss": 0.0726,
+      "step": 13525
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 8.21048703116253e-06,
+      "loss": 0.0854,
+      "step": 13526
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 8.210353744035403e-06,
+      "loss": 0.1145,
+      "step": 13527
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 8.210220456908272e-06,
+      "loss": 0.1264,
+      "step": 13528
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 8.210087169781143e-06,
+      "loss": 0.1159,
+      "step": 13529
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 8.209953882654014e-06,
+      "loss": 0.1242,
+      "step": 13530
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 8.209820595526885e-06,
+      "loss": 0.1175,
+      "step": 13531
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 8.209687308399756e-06,
+      "loss": 0.0951,
+      "step": 13532
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 8.209554021272627e-06,
+      "loss": 0.1496,
+      "step": 13533
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 8.209420734145496e-06,
+      "loss": 0.0947,
+      "step": 13534
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 8.209287447018367e-06,
+      "loss": 0.091,
+      "step": 13535
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 8.209154159891238e-06,
+      "loss": 0.1252,
+      "step": 13536
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 8.20902087276411e-06,
+      "loss": 0.1025,
+      "step": 13537
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 8.20888758563698e-06,
+      "loss": 0.0494,
+      "step": 13538
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 8.20875429850985e-06,
+      "loss": 0.1278,
+      "step": 13539
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 8.20862101138272e-06,
+      "loss": 0.0965,
+      "step": 13540
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 8.208487724255592e-06,
+      "loss": 0.1139,
+      "step": 13541
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 8.208354437128463e-06,
+      "loss": 0.0876,
+      "step": 13542
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 8.208221150001334e-06,
+      "loss": 0.1201,
+      "step": 13543
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 8.208087862874205e-06,
+      "loss": 0.1048,
+      "step": 13544
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 8.207954575747074e-06,
+      "loss": 0.1356,
+      "step": 13545
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 8.207821288619947e-06,
+      "loss": 0.0985,
+      "step": 13546
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 8.207688001492816e-06,
+      "loss": 0.1363,
+      "step": 13547
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 8.207554714365687e-06,
+      "loss": 0.1016,
+      "step": 13548
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 8.207421427238558e-06,
+      "loss": 0.0696,
+      "step": 13549
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 8.207288140111429e-06,
+      "loss": 0.0854,
+      "step": 13550
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 8.2071548529843e-06,
+      "loss": 0.119,
+      "step": 13551
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 8.20702156585717e-06,
+      "loss": 0.0767,
+      "step": 13552
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 8.20688827873004e-06,
+      "loss": 0.1069,
+      "step": 13553
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 8.206754991602913e-06,
+      "loss": 0.1303,
+      "step": 13554
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 8.206621704475782e-06,
+      "loss": 0.1131,
+      "step": 13555
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 8.206488417348653e-06,
+      "loss": 0.1129,
+      "step": 13556
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 8.206355130221524e-06,
+      "loss": 0.0828,
+      "step": 13557
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 8.206221843094393e-06,
+      "loss": 0.0846,
+      "step": 13558
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 8.206088555967266e-06,
+      "loss": 0.0914,
+      "step": 13559
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 8.205955268840135e-06,
+      "loss": 0.1028,
+      "step": 13560
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 8.205821981713006e-06,
+      "loss": 0.1028,
+      "step": 13561
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 8.205688694585877e-06,
+      "loss": 0.1094,
+      "step": 13562
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 8.205555407458748e-06,
+      "loss": 0.1213,
+      "step": 13563
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 8.20542212033162e-06,
+      "loss": 0.0779,
+      "step": 13564
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 8.20528883320449e-06,
+      "loss": 0.1019,
+      "step": 13565
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 8.20515554607736e-06,
+      "loss": 0.0854,
+      "step": 13566
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 8.205022258950232e-06,
+      "loss": 0.0898,
+      "step": 13567
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 8.204888971823102e-06,
+      "loss": 0.0849,
+      "step": 13568
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 8.204755684695973e-06,
+      "loss": 0.0815,
+      "step": 13569
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 8.204622397568844e-06,
+      "loss": 0.1273,
+      "step": 13570
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 8.204489110441715e-06,
+      "loss": 0.1257,
+      "step": 13571
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 8.204355823314586e-06,
+      "loss": 0.0983,
+      "step": 13572
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 8.204222536187457e-06,
+      "loss": 0.0795,
+      "step": 13573
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 8.204089249060326e-06,
+      "loss": 0.0859,
+      "step": 13574
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 8.203955961933197e-06,
+      "loss": 0.0867,
+      "step": 13575
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 8.203822674806068e-06,
+      "loss": 0.0966,
+      "step": 13576
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 8.203689387678939e-06,
+      "loss": 0.0944,
+      "step": 13577
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 8.20355610055181e-06,
+      "loss": 0.0604,
+      "step": 13578
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 8.20342281342468e-06,
+      "loss": 0.0604,
+      "step": 13579
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 8.20328952629755e-06,
+      "loss": 0.1154,
+      "step": 13580
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 8.203156239170421e-06,
+      "loss": 0.0737,
+      "step": 13581
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 8.203022952043292e-06,
+      "loss": 0.1426,
+      "step": 13582
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 8.202889664916163e-06,
+      "loss": 0.1359,
+      "step": 13583
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 8.202756377789034e-06,
+      "loss": 0.0982,
+      "step": 13584
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 8.202623090661903e-06,
+      "loss": 0.1275,
+      "step": 13585
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 8.202489803534776e-06,
+      "loss": 0.0792,
+      "step": 13586
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 8.202356516407645e-06,
+      "loss": 0.1552,
+      "step": 13587
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 8.202223229280516e-06,
+      "loss": 0.0839,
+      "step": 13588
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 8.202089942153387e-06,
+      "loss": 0.1232,
+      "step": 13589
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 8.201956655026258e-06,
+      "loss": 0.11,
+      "step": 13590
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 8.20182336789913e-06,
+      "loss": 0.1202,
+      "step": 13591
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 8.201690080772e-06,
+      "loss": 0.1265,
+      "step": 13592
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 8.20155679364487e-06,
+      "loss": 0.0888,
+      "step": 13593
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 8.201423506517742e-06,
+      "loss": 0.1367,
+      "step": 13594
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 8.201290219390612e-06,
+      "loss": 0.087,
+      "step": 13595
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 8.201156932263483e-06,
+      "loss": 0.1016,
+      "step": 13596
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 8.201023645136354e-06,
+      "loss": 0.1126,
+      "step": 13597
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 8.200890358009223e-06,
+      "loss": 0.092,
+      "step": 13598
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 8.200757070882096e-06,
+      "loss": 0.076,
+      "step": 13599
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 8.200623783754965e-06,
+      "loss": 0.1251,
+      "step": 13600
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 8.200490496627836e-06,
+      "loss": 0.0884,
+      "step": 13601
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 8.200357209500707e-06,
+      "loss": 0.0657,
+      "step": 13602
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 8.200223922373578e-06,
+      "loss": 0.1214,
+      "step": 13603
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 8.200090635246449e-06,
+      "loss": 0.1445,
+      "step": 13604
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 8.19995734811932e-06,
+      "loss": 0.1176,
+      "step": 13605
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 8.19982406099219e-06,
+      "loss": 0.1166,
+      "step": 13606
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 8.19969077386506e-06,
+      "loss": 0.0975,
+      "step": 13607
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 8.199557486737931e-06,
+      "loss": 0.1386,
+      "step": 13608
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 8.199424199610802e-06,
+      "loss": 0.1107,
+      "step": 13609
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 8.199290912483673e-06,
+      "loss": 0.0863,
+      "step": 13610
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 8.199157625356544e-06,
+      "loss": 0.1327,
+      "step": 13611
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 8.199024338229414e-06,
+      "loss": 0.1034,
+      "step": 13612
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 8.198891051102286e-06,
+      "loss": 0.0947,
+      "step": 13613
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 8.198757763975156e-06,
+      "loss": 0.0999,
+      "step": 13614
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 8.198624476848027e-06,
+      "loss": 0.1158,
+      "step": 13615
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 8.198491189720898e-06,
+      "loss": 0.1467,
+      "step": 13616
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 8.198357902593767e-06,
+      "loss": 0.1258,
+      "step": 13617
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 8.19822461546664e-06,
+      "loss": 0.0973,
+      "step": 13618
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 8.198091328339509e-06,
+      "loss": 0.1367,
+      "step": 13619
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 8.19795804121238e-06,
+      "loss": 0.1535,
+      "step": 13620
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 8.19782475408525e-06,
+      "loss": 0.1241,
+      "step": 13621
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 8.197691466958122e-06,
+      "loss": 0.1491,
+      "step": 13622
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 8.197558179830993e-06,
+      "loss": 0.0831,
+      "step": 13623
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 8.197424892703864e-06,
+      "loss": 0.1193,
+      "step": 13624
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 8.197291605576733e-06,
+      "loss": 0.0851,
+      "step": 13625
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 8.197158318449606e-06,
+      "loss": 0.1038,
+      "step": 13626
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 8.197025031322475e-06,
+      "loss": 0.0639,
+      "step": 13627
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 8.196891744195346e-06,
+      "loss": 0.1117,
+      "step": 13628
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 8.196758457068217e-06,
+      "loss": 0.1246,
+      "step": 13629
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 8.196625169941088e-06,
+      "loss": 0.0666,
+      "step": 13630
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 8.196491882813959e-06,
+      "loss": 0.1296,
+      "step": 13631
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 8.19635859568683e-06,
+      "loss": 0.1072,
+      "step": 13632
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 8.1962253085597e-06,
+      "loss": 0.0877,
+      "step": 13633
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 8.196092021432572e-06,
+      "loss": 0.0895,
+      "step": 13634
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 8.195958734305441e-06,
+      "loss": 0.1076,
+      "step": 13635
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 8.195825447178312e-06,
+      "loss": 0.1156,
+      "step": 13636
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 8.195692160051183e-06,
+      "loss": 0.1147,
+      "step": 13637
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 8.195558872924053e-06,
+      "loss": 0.0919,
+      "step": 13638
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 8.195425585796925e-06,
+      "loss": 0.1088,
+      "step": 13639
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 8.195292298669795e-06,
+      "loss": 0.1272,
+      "step": 13640
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 8.195159011542666e-06,
+      "loss": 0.1297,
+      "step": 13641
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 8.195025724415537e-06,
+      "loss": 0.0935,
+      "step": 13642
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 8.194892437288408e-06,
+      "loss": 0.103,
+      "step": 13643
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 8.194759150161279e-06,
+      "loss": 0.1224,
+      "step": 13644
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 8.19462586303415e-06,
+      "loss": 0.0884,
+      "step": 13645
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 8.194492575907019e-06,
+      "loss": 0.1003,
+      "step": 13646
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 8.19435928877989e-06,
+      "loss": 0.1227,
+      "step": 13647
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 8.194226001652761e-06,
+      "loss": 0.1021,
+      "step": 13648
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.194092714525632e-06,
+      "loss": 0.0948,
+      "step": 13649
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.193959427398503e-06,
+      "loss": 0.1088,
+      "step": 13650
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.193826140271374e-06,
+      "loss": 0.0918,
+      "step": 13651
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.193692853144243e-06,
+      "loss": 0.1089,
+      "step": 13652
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.193559566017116e-06,
+      "loss": 0.0958,
+      "step": 13653
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.193426278889985e-06,
+      "loss": 0.0751,
+      "step": 13654
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.193292991762856e-06,
+      "loss": 0.1399,
+      "step": 13655
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.193159704635727e-06,
+      "loss": 0.1015,
+      "step": 13656
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.193026417508596e-06,
+      "loss": 0.0944,
+      "step": 13657
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.192893130381469e-06,
+      "loss": 0.0649,
+      "step": 13658
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.192759843254338e-06,
+      "loss": 0.1469,
+      "step": 13659
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.19262655612721e-06,
+      "loss": 0.1074,
+      "step": 13660
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.19249326900008e-06,
+      "loss": 0.1079,
+      "step": 13661
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.192359981872951e-06,
+      "loss": 0.1052,
+      "step": 13662
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.192226694745822e-06,
+      "loss": 0.0766,
+      "step": 13663
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.192093407618693e-06,
+      "loss": 0.0796,
+      "step": 13664
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.191960120491563e-06,
+      "loss": 0.1153,
+      "step": 13665
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.191826833364435e-06,
+      "loss": 0.1226,
+      "step": 13666
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.191693546237305e-06,
+      "loss": 0.0997,
+      "step": 13667
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.191560259110176e-06,
+      "loss": 0.0956,
+      "step": 13668
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.191426971983047e-06,
+      "loss": 0.1256,
+      "step": 13669
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.191293684855918e-06,
+      "loss": 0.105,
+      "step": 13670
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.191160397728789e-06,
+      "loss": 0.1082,
+      "step": 13671
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.19102711060166e-06,
+      "loss": 0.1192,
+      "step": 13672
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.190893823474529e-06,
+      "loss": 0.0935,
+      "step": 13673
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.1907605363474e-06,
+      "loss": 0.0768,
+      "step": 13674
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.190627249220271e-06,
+      "loss": 0.0886,
+      "step": 13675
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.190493962093142e-06,
+      "loss": 0.1056,
+      "step": 13676
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.190360674966013e-06,
+      "loss": 0.086,
+      "step": 13677
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.190227387838882e-06,
+      "loss": 0.0695,
+      "step": 13678
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.190094100711753e-06,
+      "loss": 0.1015,
+      "step": 13679
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.189960813584624e-06,
+      "loss": 0.0784,
+      "step": 13680
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.189827526457495e-06,
+      "loss": 0.0906,
+      "step": 13681
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.189694239330366e-06,
+      "loss": 0.1281,
+      "step": 13682
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.189560952203237e-06,
+      "loss": 0.0893,
+      "step": 13683
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.189427665076107e-06,
+      "loss": 0.1282,
+      "step": 13684
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.18929437794898e-06,
+      "loss": 0.0629,
+      "step": 13685
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.189161090821849e-06,
+      "loss": 0.0871,
+      "step": 13686
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.18902780369472e-06,
+      "loss": 0.1009,
+      "step": 13687
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.18889451656759e-06,
+      "loss": 0.1048,
+      "step": 13688
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.188761229440461e-06,
+      "loss": 0.0953,
+      "step": 13689
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.188627942313332e-06,
+      "loss": 0.1002,
+      "step": 13690
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.188494655186203e-06,
+      "loss": 0.106,
+      "step": 13691
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.188361368059073e-06,
+      "loss": 0.1111,
+      "step": 13692
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.188228080931945e-06,
+      "loss": 0.1056,
+      "step": 13693
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.188094793804815e-06,
+      "loss": 0.0965,
+      "step": 13694
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.187961506677686e-06,
+      "loss": 0.0619,
+      "step": 13695
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.187828219550557e-06,
+      "loss": 0.0918,
+      "step": 13696
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.187694932423426e-06,
+      "loss": 0.1455,
+      "step": 13697
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.187561645296299e-06,
+      "loss": 0.12,
+      "step": 13698
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.187428358169168e-06,
+      "loss": 0.1034,
+      "step": 13699
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.187295071042039e-06,
+      "loss": 0.0937,
+      "step": 13700
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.18716178391491e-06,
+      "loss": 0.0994,
+      "step": 13701
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.187028496787781e-06,
+      "loss": 0.0924,
+      "step": 13702
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.186895209660652e-06,
+      "loss": 0.0861,
+      "step": 13703
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.186761922533523e-06,
+      "loss": 0.1034,
+      "step": 13704
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.186628635406392e-06,
+      "loss": 0.1309,
+      "step": 13705
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.186495348279265e-06,
+      "loss": 0.1215,
+      "step": 13706
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.186362061152134e-06,
+      "loss": 0.115,
+      "step": 13707
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.186228774025005e-06,
+      "loss": 0.0978,
+      "step": 13708
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.186095486897876e-06,
+      "loss": 0.1147,
+      "step": 13709
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.185962199770747e-06,
+      "loss": 0.1133,
+      "step": 13710
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.185828912643618e-06,
+      "loss": 0.1504,
+      "step": 13711
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.18569562551649e-06,
+      "loss": 0.0866,
+      "step": 13712
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.185562338389359e-06,
+      "loss": 0.0912,
+      "step": 13713
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.18542905126223e-06,
+      "loss": 0.1474,
+      "step": 13714
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.1852957641351e-06,
+      "loss": 0.0787,
+      "step": 13715
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.18516247700797e-06,
+      "loss": 0.0868,
+      "step": 13716
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.185029189880843e-06,
+      "loss": 0.1471,
+      "step": 13717
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.184895902753712e-06,
+      "loss": 0.0878,
+      "step": 13718
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.184762615626583e-06,
+      "loss": 0.0894,
+      "step": 13719
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.184629328499454e-06,
+      "loss": 0.066,
+      "step": 13720
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.184496041372325e-06,
+      "loss": 0.1366,
+      "step": 13721
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.184362754245196e-06,
+      "loss": 0.0957,
+      "step": 13722
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.184229467118067e-06,
+      "loss": 0.103,
+      "step": 13723
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.184096179990936e-06,
+      "loss": 0.1355,
+      "step": 13724
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.183962892863809e-06,
+      "loss": 0.1106,
+      "step": 13725
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.183829605736678e-06,
+      "loss": 0.1254,
+      "step": 13726
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.183696318609549e-06,
+      "loss": 0.1207,
+      "step": 13727
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.18356303148242e-06,
+      "loss": 0.1004,
+      "step": 13728
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.183429744355291e-06,
+      "loss": 0.0726,
+      "step": 13729
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.183296457228162e-06,
+      "loss": 0.1329,
+      "step": 13730
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.183163170101033e-06,
+      "loss": 0.1356,
+      "step": 13731
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.183029882973902e-06,
+      "loss": 0.0934,
+      "step": 13732
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.182896595846775e-06,
+      "loss": 0.1207,
+      "step": 13733
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.182763308719644e-06,
+      "loss": 0.0916,
+      "step": 13734
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.182630021592515e-06,
+      "loss": 0.0764,
+      "step": 13735
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.182496734465386e-06,
+      "loss": 0.1217,
+      "step": 13736
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.182363447338256e-06,
+      "loss": 0.1163,
+      "step": 13737
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.182230160211128e-06,
+      "loss": 0.0894,
+      "step": 13738
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.182096873083998e-06,
+      "loss": 0.0877,
+      "step": 13739
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.181963585956869e-06,
+      "loss": 0.111,
+      "step": 13740
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.18183029882974e-06,
+      "loss": 0.1053,
+      "step": 13741
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.18169701170261e-06,
+      "loss": 0.095,
+      "step": 13742
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.181563724575482e-06,
+      "loss": 0.0941,
+      "step": 13743
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.181430437448353e-06,
+      "loss": 0.1088,
+      "step": 13744
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.181297150321222e-06,
+      "loss": 0.0948,
+      "step": 13745
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.181163863194093e-06,
+      "loss": 0.0675,
+      "step": 13746
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.181030576066964e-06,
+      "loss": 0.1521,
+      "step": 13747
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.180897288939835e-06,
+      "loss": 0.094,
+      "step": 13748
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.180764001812706e-06,
+      "loss": 0.0887,
+      "step": 13749
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.180630714685577e-06,
+      "loss": 0.0785,
+      "step": 13750
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.180497427558446e-06,
+      "loss": 0.076,
+      "step": 13751
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.180364140431319e-06,
+      "loss": 0.1333,
+      "step": 13752
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.180230853304188e-06,
+      "loss": 0.1001,
+      "step": 13753
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.18009756617706e-06,
+      "loss": 0.0824,
+      "step": 13754
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.17996427904993e-06,
+      "loss": 0.0946,
+      "step": 13755
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.1798309919228e-06,
+      "loss": 0.1276,
+      "step": 13756
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.179697704795672e-06,
+      "loss": 0.0853,
+      "step": 13757
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.179564417668541e-06,
+      "loss": 0.1426,
+      "step": 13758
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.179431130541412e-06,
+      "loss": 0.123,
+      "step": 13759
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.179297843414283e-06,
+      "loss": 0.1316,
+      "step": 13760
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.179164556287154e-06,
+      "loss": 0.1164,
+      "step": 13761
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.179031269160025e-06,
+      "loss": 0.0987,
+      "step": 13762
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.178897982032896e-06,
+      "loss": 0.1082,
+      "step": 13763
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.178764694905766e-06,
+      "loss": 0.1202,
+      "step": 13764
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.178631407778638e-06,
+      "loss": 0.1027,
+      "step": 13765
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.178498120651508e-06,
+      "loss": 0.0948,
+      "step": 13766
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.178364833524379e-06,
+      "loss": 0.0807,
+      "step": 13767
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.17823154639725e-06,
+      "loss": 0.1046,
+      "step": 13768
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.17809825927012e-06,
+      "loss": 0.1289,
+      "step": 13769
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.177964972142992e-06,
+      "loss": 0.1237,
+      "step": 13770
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.177831685015863e-06,
+      "loss": 0.1595,
+      "step": 13771
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.177698397888732e-06,
+      "loss": 0.086,
+      "step": 13772
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.177565110761603e-06,
+      "loss": 0.1541,
+      "step": 13773
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.177431823634474e-06,
+      "loss": 0.0783,
+      "step": 13774
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.177298536507345e-06,
+      "loss": 0.1381,
+      "step": 13775
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.177165249380216e-06,
+      "loss": 0.0971,
+      "step": 13776
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.177031962253085e-06,
+      "loss": 0.0973,
+      "step": 13777
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.176898675125956e-06,
+      "loss": 0.1084,
+      "step": 13778
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.176765387998827e-06,
+      "loss": 0.1541,
+      "step": 13779
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.176632100871698e-06,
+      "loss": 0.1039,
+      "step": 13780
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.17649881374457e-06,
+      "loss": 0.0637,
+      "step": 13781
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.17636552661744e-06,
+      "loss": 0.1179,
+      "step": 13782
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.17623223949031e-06,
+      "loss": 0.095,
+      "step": 13783
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.176098952363182e-06,
+      "loss": 0.1339,
+      "step": 13784
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.175965665236052e-06,
+      "loss": 0.1082,
+      "step": 13785
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.175832378108923e-06,
+      "loss": 0.1048,
+      "step": 13786
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.175699090981794e-06,
+      "loss": 0.0954,
+      "step": 13787
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.175565803854665e-06,
+      "loss": 0.1146,
+      "step": 13788
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.175432516727536e-06,
+      "loss": 0.0565,
+      "step": 13789
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.175299229600407e-06,
+      "loss": 0.1066,
+      "step": 13790
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.175165942473276e-06,
+      "loss": 0.1253,
+      "step": 13791
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.175032655346149e-06,
+      "loss": 0.1246,
+      "step": 13792
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.174899368219018e-06,
+      "loss": 0.0663,
+      "step": 13793
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.174766081091889e-06,
+      "loss": 0.1175,
+      "step": 13794
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.17463279396476e-06,
+      "loss": 0.1005,
+      "step": 13795
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.174499506837629e-06,
+      "loss": 0.123,
+      "step": 13796
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.174366219710502e-06,
+      "loss": 0.0917,
+      "step": 13797
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.174232932583371e-06,
+      "loss": 0.0721,
+      "step": 13798
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.174099645456242e-06,
+      "loss": 0.1,
+      "step": 13799
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.173966358329113e-06,
+      "loss": 0.1468,
+      "step": 13800
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.173833071201984e-06,
+      "loss": 0.1295,
+      "step": 13801
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.173699784074855e-06,
+      "loss": 0.0864,
+      "step": 13802
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.173566496947726e-06,
+      "loss": 0.1112,
+      "step": 13803
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.173433209820595e-06,
+      "loss": 0.0854,
+      "step": 13804
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.173299922693468e-06,
+      "loss": 0.1245,
+      "step": 13805
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.173166635566337e-06,
+      "loss": 0.0848,
+      "step": 13806
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.173033348439208e-06,
+      "loss": 0.0735,
+      "step": 13807
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.17290006131208e-06,
+      "loss": 0.0909,
+      "step": 13808
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.17276677418495e-06,
+      "loss": 0.0871,
+      "step": 13809
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.172633487057821e-06,
+      "loss": 0.1361,
+      "step": 13810
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.172500199930692e-06,
+      "loss": 0.1342,
+      "step": 13811
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.172366912803562e-06,
+      "loss": 0.1008,
+      "step": 13812
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.172233625676433e-06,
+      "loss": 0.1318,
+      "step": 13813
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.172100338549304e-06,
+      "loss": 0.1281,
+      "step": 13814
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.171967051422175e-06,
+      "loss": 0.1062,
+      "step": 13815
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.171833764295046e-06,
+      "loss": 0.0775,
+      "step": 13816
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.171700477167915e-06,
+      "loss": 0.1204,
+      "step": 13817
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.171567190040786e-06,
+      "loss": 0.1545,
+      "step": 13818
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.171433902913657e-06,
+      "loss": 0.1522,
+      "step": 13819
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.171300615786528e-06,
+      "loss": 0.1185,
+      "step": 13820
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.171167328659399e-06,
+      "loss": 0.0998,
+      "step": 13821
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.17103404153227e-06,
+      "loss": 0.1088,
+      "step": 13822
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.17090075440514e-06,
+      "loss": 0.066,
+      "step": 13823
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.170767467278012e-06,
+      "loss": 0.0952,
+      "step": 13824
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.170634180150881e-06,
+      "loss": 0.0769,
+      "step": 13825
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.170500893023752e-06,
+      "loss": 0.0858,
+      "step": 13826
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.170367605896623e-06,
+      "loss": 0.1011,
+      "step": 13827
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.170234318769494e-06,
+      "loss": 0.0892,
+      "step": 13828
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.170101031642365e-06,
+      "loss": 0.1241,
+      "step": 13829
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.169967744515236e-06,
+      "loss": 0.1173,
+      "step": 13830
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.169834457388105e-06,
+      "loss": 0.105,
+      "step": 13831
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.169701170260976e-06,
+      "loss": 0.0739,
+      "step": 13832
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.169567883133847e-06,
+      "loss": 0.0828,
+      "step": 13833
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.169434596006718e-06,
+      "loss": 0.0955,
+      "step": 13834
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.16930130887959e-06,
+      "loss": 0.0981,
+      "step": 13835
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.169168021752459e-06,
+      "loss": 0.0926,
+      "step": 13836
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.169034734625331e-06,
+      "loss": 0.1519,
+      "step": 13837
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.1689014474982e-06,
+      "loss": 0.1056,
+      "step": 13838
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.168768160371072e-06,
+      "loss": 0.1083,
+      "step": 13839
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.168634873243943e-06,
+      "loss": 0.0772,
+      "step": 13840
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.168501586116814e-06,
+      "loss": 0.0887,
+      "step": 13841
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.168368298989685e-06,
+      "loss": 0.0946,
+      "step": 13842
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.168235011862556e-06,
+      "loss": 0.0941,
+      "step": 13843
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.168101724735425e-06,
+      "loss": 0.1377,
+      "step": 13844
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.167968437608296e-06,
+      "loss": 0.0989,
+      "step": 13845
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.167835150481167e-06,
+      "loss": 0.0784,
+      "step": 13846
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.167701863354038e-06,
+      "loss": 0.1063,
+      "step": 13847
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.167568576226909e-06,
+      "loss": 0.0987,
+      "step": 13848
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.16743528909978e-06,
+      "loss": 0.1406,
+      "step": 13849
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.16730200197265e-06,
+      "loss": 0.1211,
+      "step": 13850
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.167168714845522e-06,
+      "loss": 0.0803,
+      "step": 13851
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.167035427718391e-06,
+      "loss": 0.091,
+      "step": 13852
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.166902140591262e-06,
+      "loss": 0.0933,
+      "step": 13853
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.166768853464133e-06,
+      "loss": 0.1039,
+      "step": 13854
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.166635566337003e-06,
+      "loss": 0.0894,
+      "step": 13855
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.166502279209875e-06,
+      "loss": 0.086,
+      "step": 13856
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.166368992082745e-06,
+      "loss": 0.1183,
+      "step": 13857
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.166235704955616e-06,
+      "loss": 0.1347,
+      "step": 13858
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.166102417828487e-06,
+      "loss": 0.1125,
+      "step": 13859
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.165969130701358e-06,
+      "loss": 0.1244,
+      "step": 13860
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.165835843574229e-06,
+      "loss": 0.0836,
+      "step": 13861
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.1657025564471e-06,
+      "loss": 0.0888,
+      "step": 13862
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.165569269319969e-06,
+      "loss": 0.1235,
+      "step": 13863
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.165435982192842e-06,
+      "loss": 0.1232,
+      "step": 13864
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.16530269506571e-06,
+      "loss": 0.0963,
+      "step": 13865
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.165169407938582e-06,
+      "loss": 0.0755,
+      "step": 13866
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.165036120811453e-06,
+      "loss": 0.122,
+      "step": 13867
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.164902833684324e-06,
+      "loss": 0.0975,
+      "step": 13868
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.164769546557195e-06,
+      "loss": 0.0859,
+      "step": 13869
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.164636259430066e-06,
+      "loss": 0.0896,
+      "step": 13870
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.164502972302935e-06,
+      "loss": 0.1059,
+      "step": 13871
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.164369685175806e-06,
+      "loss": 0.0735,
+      "step": 13872
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.164236398048677e-06,
+      "loss": 0.106,
+      "step": 13873
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.164103110921548e-06,
+      "loss": 0.1262,
+      "step": 13874
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.163969823794419e-06,
+      "loss": 0.1446,
+      "step": 13875
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.163836536667288e-06,
+      "loss": 0.1287,
+      "step": 13876
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.163703249540161e-06,
+      "loss": 0.0741,
+      "step": 13877
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.16356996241303e-06,
+      "loss": 0.1231,
+      "step": 13878
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.163436675285901e-06,
+      "loss": 0.1368,
+      "step": 13879
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.163303388158772e-06,
+      "loss": 0.0939,
+      "step": 13880
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.163170101031643e-06,
+      "loss": 0.1723,
+      "step": 13881
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.163036813904514e-06,
+      "loss": 0.1049,
+      "step": 13882
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.162903526777385e-06,
+      "loss": 0.0852,
+      "step": 13883
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.162770239650255e-06,
+      "loss": 0.1099,
+      "step": 13884
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.162636952523126e-06,
+      "loss": 0.1096,
+      "step": 13885
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.162503665395997e-06,
+      "loss": 0.1108,
+      "step": 13886
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.162370378268868e-06,
+      "loss": 0.1144,
+      "step": 13887
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.162237091141739e-06,
+      "loss": 0.0804,
+      "step": 13888
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.16210380401461e-06,
+      "loss": 0.1162,
+      "step": 13889
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.161970516887479e-06,
+      "loss": 0.1031,
+      "step": 13890
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.16183722976035e-06,
+      "loss": 0.1163,
+      "step": 13891
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.161703942633221e-06,
+      "loss": 0.1049,
+      "step": 13892
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.161570655506092e-06,
+      "loss": 0.0852,
+      "step": 13893
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.161437368378963e-06,
+      "loss": 0.0954,
+      "step": 13894
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.161304081251832e-06,
+      "loss": 0.091,
+      "step": 13895
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.161170794124705e-06,
+      "loss": 0.0974,
+      "step": 13896
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.161037506997574e-06,
+      "loss": 0.0991,
+      "step": 13897
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.160904219870445e-06,
+      "loss": 0.1631,
+      "step": 13898
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.160770932743316e-06,
+      "loss": 0.0668,
+      "step": 13899
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.160637645616187e-06,
+      "loss": 0.0754,
+      "step": 13900
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.160504358489058e-06,
+      "loss": 0.1137,
+      "step": 13901
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.16037107136193e-06,
+      "loss": 0.0952,
+      "step": 13902
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.160237784234798e-06,
+      "loss": 0.0774,
+      "step": 13903
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.160104497107671e-06,
+      "loss": 0.0993,
+      "step": 13904
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.15997120998054e-06,
+      "loss": 0.1312,
+      "step": 13905
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.159837922853411e-06,
+      "loss": 0.1182,
+      "step": 13906
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.159704635726282e-06,
+      "loss": 0.0641,
+      "step": 13907
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.159571348599153e-06,
+      "loss": 0.0825,
+      "step": 13908
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.159438061472024e-06,
+      "loss": 0.0879,
+      "step": 13909
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.159304774344895e-06,
+      "loss": 0.1068,
+      "step": 13910
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.159171487217765e-06,
+      "loss": 0.0796,
+      "step": 13911
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.159038200090636e-06,
+      "loss": 0.1417,
+      "step": 13912
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.158904912963507e-06,
+      "loss": 0.1069,
+      "step": 13913
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.158771625836378e-06,
+      "loss": 0.0813,
+      "step": 13914
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.158638338709249e-06,
+      "loss": 0.0863,
+      "step": 13915
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.158505051582118e-06,
+      "loss": 0.0694,
+      "step": 13916
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.158371764454989e-06,
+      "loss": 0.1595,
+      "step": 13917
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.15823847732786e-06,
+      "loss": 0.1161,
+      "step": 13918
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.158105190200731e-06,
+      "loss": 0.1136,
+      "step": 13919
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.157971903073602e-06,
+      "loss": 0.0838,
+      "step": 13920
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.157838615946473e-06,
+      "loss": 0.088,
+      "step": 13921
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.157705328819342e-06,
+      "loss": 0.0835,
+      "step": 13922
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.157572041692215e-06,
+      "loss": 0.1099,
+      "step": 13923
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.157438754565084e-06,
+      "loss": 0.1091,
+      "step": 13924
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.157305467437955e-06,
+      "loss": 0.1025,
+      "step": 13925
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.157172180310826e-06,
+      "loss": 0.1325,
+      "step": 13926
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.157038893183697e-06,
+      "loss": 0.112,
+      "step": 13927
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.156905606056568e-06,
+      "loss": 0.0871,
+      "step": 13928
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.15677231892944e-06,
+      "loss": 0.0974,
+      "step": 13929
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.156639031802309e-06,
+      "loss": 0.0762,
+      "step": 13930
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.15650574467518e-06,
+      "loss": 0.1245,
+      "step": 13931
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.15637245754805e-06,
+      "loss": 0.1305,
+      "step": 13932
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.156239170420922e-06,
+      "loss": 0.1493,
+      "step": 13933
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.156105883293793e-06,
+      "loss": 0.0812,
+      "step": 13934
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.155972596166662e-06,
+      "loss": 0.1052,
+      "step": 13935
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.155839309039535e-06,
+      "loss": 0.1252,
+      "step": 13936
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.155706021912404e-06,
+      "loss": 0.0932,
+      "step": 13937
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.155572734785275e-06,
+      "loss": 0.121,
+      "step": 13938
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.155439447658146e-06,
+      "loss": 0.1039,
+      "step": 13939
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.155306160531017e-06,
+      "loss": 0.0784,
+      "step": 13940
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.155172873403888e-06,
+      "loss": 0.093,
+      "step": 13941
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.155039586276759e-06,
+      "loss": 0.1742,
+      "step": 13942
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.154906299149628e-06,
+      "loss": 0.093,
+      "step": 13943
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.1547730120225e-06,
+      "loss": 0.0939,
+      "step": 13944
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.15463972489537e-06,
+      "loss": 0.0817,
+      "step": 13945
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.154506437768241e-06,
+      "loss": 0.1152,
+      "step": 13946
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.154373150641112e-06,
+      "loss": 0.1038,
+      "step": 13947
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.154239863513983e-06,
+      "loss": 0.086,
+      "step": 13948
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.154106576386854e-06,
+      "loss": 0.1504,
+      "step": 13949
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.153973289259723e-06,
+      "loss": 0.1144,
+      "step": 13950
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.153840002132594e-06,
+      "loss": 0.1236,
+      "step": 13951
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.153706715005465e-06,
+      "loss": 0.1168,
+      "step": 13952
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.153573427878336e-06,
+      "loss": 0.1367,
+      "step": 13953
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.153440140751207e-06,
+      "loss": 0.1213,
+      "step": 13954
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.153306853624078e-06,
+      "loss": 0.1064,
+      "step": 13955
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.153173566496948e-06,
+      "loss": 0.104,
+      "step": 13956
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.153040279369819e-06,
+      "loss": 0.098,
+      "step": 13957
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.15290699224269e-06,
+      "loss": 0.1185,
+      "step": 13958
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.15277370511556e-06,
+      "loss": 0.1033,
+      "step": 13959
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.152640417988432e-06,
+      "loss": 0.1122,
+      "step": 13960
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.152507130861303e-06,
+      "loss": 0.099,
+      "step": 13961
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.152373843734172e-06,
+      "loss": 0.1088,
+      "step": 13962
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.152240556607045e-06,
+      "loss": 0.0841,
+      "step": 13963
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.152107269479914e-06,
+      "loss": 0.1143,
+      "step": 13964
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.151973982352785e-06,
+      "loss": 0.0676,
+      "step": 13965
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.151840695225656e-06,
+      "loss": 0.1127,
+      "step": 13966
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.151707408098527e-06,
+      "loss": 0.1172,
+      "step": 13967
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.151574120971398e-06,
+      "loss": 0.0698,
+      "step": 13968
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.151440833844269e-06,
+      "loss": 0.1216,
+      "step": 13969
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.151307546717138e-06,
+      "loss": 0.1035,
+      "step": 13970
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.15117425959001e-06,
+      "loss": 0.116,
+      "step": 13971
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.15104097246288e-06,
+      "loss": 0.1187,
+      "step": 13972
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.150907685335751e-06,
+      "loss": 0.1063,
+      "step": 13973
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.150774398208622e-06,
+      "loss": 0.1265,
+      "step": 13974
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.150641111081491e-06,
+      "loss": 0.1044,
+      "step": 13975
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.150507823954364e-06,
+      "loss": 0.1083,
+      "step": 13976
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.150374536827233e-06,
+      "loss": 0.09,
+      "step": 13977
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.150241249700104e-06,
+      "loss": 0.124,
+      "step": 13978
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.150107962572975e-06,
+      "loss": 0.0796,
+      "step": 13979
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.149974675445846e-06,
+      "loss": 0.0952,
+      "step": 13980
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.149841388318717e-06,
+      "loss": 0.1412,
+      "step": 13981
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.149708101191588e-06,
+      "loss": 0.0871,
+      "step": 13982
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.149574814064458e-06,
+      "loss": 0.0699,
+      "step": 13983
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.149441526937329e-06,
+      "loss": 0.0912,
+      "step": 13984
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.1493082398102e-06,
+      "loss": 0.0949,
+      "step": 13985
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.14917495268307e-06,
+      "loss": 0.1235,
+      "step": 13986
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.149041665555942e-06,
+      "loss": 0.0774,
+      "step": 13987
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.148908378428813e-06,
+      "loss": 0.0981,
+      "step": 13988
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.148775091301682e-06,
+      "loss": 0.072,
+      "step": 13989
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.148641804174553e-06,
+      "loss": 0.1167,
+      "step": 13990
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.148508517047424e-06,
+      "loss": 0.127,
+      "step": 13991
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.148375229920295e-06,
+      "loss": 0.1094,
+      "step": 13992
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.148241942793166e-06,
+      "loss": 0.1259,
+      "step": 13993
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.148108655666035e-06,
+      "loss": 0.094,
+      "step": 13994
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.147975368538908e-06,
+      "loss": 0.0984,
+      "step": 13995
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.147842081411777e-06,
+      "loss": 0.0936,
+      "step": 13996
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.147708794284648e-06,
+      "loss": 0.0737,
+      "step": 13997
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.14757550715752e-06,
+      "loss": 0.143,
+      "step": 13998
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.14744222003039e-06,
+      "loss": 0.1006,
+      "step": 13999
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.147308932903261e-06,
+      "loss": 0.1413,
+      "step": 14000
+    }
+  ],
+  "max_steps": 75126,
+  "num_train_epochs": 3,
+  "total_flos": 1.7894879410186027e+19,
+  "trial_name": null,
+  "trial_params": null
+}